]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - kernel/trace/trace.c
ring-buffer: make ring_buffer_read_page read from start on partial page
[mirror_ubuntu-jammy-kernel.git] / kernel / trace / trace.c
CommitLineData
bc0c38d1
SR
1/*
2 * ring buffer based function tracer
3 *
4 * Copyright (C) 2007-2008 Steven Rostedt <srostedt@redhat.com>
5 * Copyright (C) 2008 Ingo Molnar <mingo@redhat.com>
6 *
7 * Originally taken from the RT patch by:
8 * Arnaldo Carvalho de Melo <acme@redhat.com>
9 *
10 * Based on code from the latency_tracer, that is:
11 * Copyright (C) 2004-2006 Ingo Molnar
12 * Copyright (C) 2004 William Lee Irwin III
13 */
14#include <linux/utsrelease.h>
15#include <linux/kallsyms.h>
16#include <linux/seq_file.h>
3f5a54e3 17#include <linux/notifier.h>
bc0c38d1 18#include <linux/debugfs.h>
4c11d7ae 19#include <linux/pagemap.h>
bc0c38d1
SR
20#include <linux/hardirq.h>
21#include <linux/linkage.h>
22#include <linux/uaccess.h>
23#include <linux/ftrace.h>
24#include <linux/module.h>
25#include <linux/percpu.h>
3f5a54e3 26#include <linux/kdebug.h>
bc0c38d1
SR
27#include <linux/ctype.h>
28#include <linux/init.h>
2a2cc8f7 29#include <linux/poll.h>
bc0c38d1
SR
30#include <linux/gfp.h>
31#include <linux/fs.h>
76094a2c 32#include <linux/kprobes.h>
3eefae99 33#include <linux/writeback.h>
3c56819b 34#include <linux/splice.h>
bc0c38d1 35
86387f7e 36#include <linux/stacktrace.h>
3928a8a2 37#include <linux/ring_buffer.h>
21798a84 38#include <linux/irqflags.h>
86387f7e 39
bc0c38d1 40#include "trace.h"
f0868d1e 41#include "trace_output.h"
bc0c38d1 42
3928a8a2
SR
43#define TRACE_BUFFER_FLAGS (RB_FL_OVERWRITE)
44
745b1626 45unsigned long __read_mostly tracing_max_latency;
bc0c38d1
SR
46unsigned long __read_mostly tracing_thresh;
47
8e1b82e0
FW
48/*
49 * We need to change this state when a selftest is running.
ff32504f
FW
50 * A selftest will lurk into the ring-buffer to count the
51 * entries inserted during the selftest although some concurrent
52 * insertions into the ring-buffer such as ftrace_printk could occurred
53 * at the same time, giving false positive or negative results.
54 */
8e1b82e0 55static bool __read_mostly tracing_selftest_running;
ff32504f 56
b2821ae6
SR
57/*
58 * If a tracer is running, we do not want to run SELFTEST.
59 */
60static bool __read_mostly tracing_selftest_disabled;
61
adf9f195
FW
62/* For tracers that don't implement custom flags */
63static struct tracer_opt dummy_tracer_opt[] = {
64 { }
65};
66
67static struct tracer_flags dummy_tracer_flags = {
68 .val = 0,
69 .opts = dummy_tracer_opt
70};
71
72static int dummy_set_flag(u32 old_flags, u32 bit, int set)
73{
74 return 0;
75}
0f048701
SR
76
77/*
78 * Kill all tracing for good (never come back).
79 * It is initialized to 1 but will turn to zero if the initialization
80 * of the tracer is successful. But that is the only place that sets
81 * this back to zero.
82 */
4fd27358 83static int tracing_disabled = 1;
0f048701 84
d769041f
SR
85static DEFINE_PER_CPU(local_t, ftrace_cpu_disabled);
86
87static inline void ftrace_disable_cpu(void)
88{
89 preempt_disable();
90 local_inc(&__get_cpu_var(ftrace_cpu_disabled));
91}
92
93static inline void ftrace_enable_cpu(void)
94{
95 local_dec(&__get_cpu_var(ftrace_cpu_disabled));
96 preempt_enable();
97}
98
9e01c1b7 99static cpumask_var_t __read_mostly tracing_buffer_mask;
ab46428c 100
b04cc6b1
FW
101/* Define which cpu buffers are currently read in trace_pipe */
102static cpumask_var_t tracing_reader_cpumask;
103
ab46428c 104#define for_each_tracing_cpu(cpu) \
9e01c1b7 105 for_each_cpu(cpu, tracing_buffer_mask)
ab46428c 106
944ac425
SR
107/*
108 * ftrace_dump_on_oops - variable to dump ftrace buffer on oops
109 *
110 * If there is an oops (or kernel panic) and the ftrace_dump_on_oops
111 * is set, then ftrace_dump is called. This will output the contents
112 * of the ftrace buffers to the console. This is very useful for
113 * capturing traces that lead to crashes and outputing it to a
114 * serial console.
115 *
116 * It is default off, but you can enable it with either specifying
117 * "ftrace_dump_on_oops" in the kernel command line, or setting
118 * /proc/sys/kernel/ftrace_dump_on_oops to true.
119 */
120int ftrace_dump_on_oops;
121
b2821ae6
SR
122static int tracing_set_tracer(const char *buf);
123
124#define BOOTUP_TRACER_SIZE 100
125static char bootup_tracer_buf[BOOTUP_TRACER_SIZE] __initdata;
126static char *default_bootup_tracer;
d9e54076
PZ
127
128static int __init set_ftrace(char *str)
129{
b2821ae6
SR
130 strncpy(bootup_tracer_buf, str, BOOTUP_TRACER_SIZE);
131 default_bootup_tracer = bootup_tracer_buf;
d9e54076
PZ
132 return 1;
133}
b2821ae6 134__setup("ftrace=", set_ftrace);
d9e54076 135
944ac425
SR
136static int __init set_ftrace_dump_on_oops(char *str)
137{
138 ftrace_dump_on_oops = 1;
139 return 1;
140}
141__setup("ftrace_dump_on_oops", set_ftrace_dump_on_oops);
60a11774 142
72829bc3 143long
bc0c38d1
SR
144ns2usecs(cycle_t nsec)
145{
146 nsec += 500;
147 do_div(nsec, 1000);
148 return nsec;
149}
150
e309b41d 151cycle_t ftrace_now(int cpu)
750ed1a4 152{
3928a8a2
SR
153 u64 ts = ring_buffer_time_stamp(cpu);
154 ring_buffer_normalize_time_stamp(cpu, &ts);
155 return ts;
750ed1a4
IM
156}
157
4fcdae83
SR
158/*
159 * The global_trace is the descriptor that holds the tracing
160 * buffers for the live tracing. For each CPU, it contains
161 * a link list of pages that will store trace entries. The
162 * page descriptor of the pages in the memory is used to hold
163 * the link list by linking the lru item in the page descriptor
164 * to each of the pages in the buffer per CPU.
165 *
166 * For each active CPU there is a data field that holds the
167 * pages for the buffer for that CPU. Each CPU has the same number
168 * of pages allocated for its buffer.
169 */
bc0c38d1
SR
170static struct trace_array global_trace;
171
172static DEFINE_PER_CPU(struct trace_array_cpu, global_trace_cpu);
173
4fcdae83
SR
174/*
175 * The max_tr is used to snapshot the global_trace when a maximum
176 * latency is reached. Some tracers will use this to store a maximum
177 * trace while it continues examining live traces.
178 *
179 * The buffers for the max_tr are set up the same as the global_trace.
180 * When a snapshot is taken, the link list of the max_tr is swapped
181 * with the link list of the global_trace and the buffers are reset for
182 * the global_trace so the tracing can continue.
183 */
bc0c38d1
SR
184static struct trace_array max_tr;
185
186static DEFINE_PER_CPU(struct trace_array_cpu, max_data);
187
4fcdae83 188/* tracer_enabled is used to toggle activation of a tracer */
26994ead 189static int tracer_enabled = 1;
4fcdae83 190
9036990d
SR
191/**
192 * tracing_is_enabled - return tracer_enabled status
193 *
194 * This function is used by other tracers to know the status
195 * of the tracer_enabled flag. Tracers may use this function
196 * to know if it should enable their features when starting
197 * up. See irqsoff tracer for an example (start_irqsoff_tracer).
198 */
199int tracing_is_enabled(void)
200{
201 return tracer_enabled;
202}
203
4fcdae83 204/*
3928a8a2
SR
205 * trace_buf_size is the size in bytes that is allocated
206 * for a buffer. Note, the number of bytes is always rounded
207 * to page size.
3f5a54e3
SR
208 *
209 * This number is purposely set to a low number of 16384.
210 * If the dump on oops happens, it will be much appreciated
211 * to not have to wait for all that output. Anyway this can be
212 * boot time and run time configurable.
4fcdae83 213 */
3928a8a2 214#define TRACE_BUF_SIZE_DEFAULT 1441792UL /* 16384 * 88 (sizeof(entry)) */
3f5a54e3 215
3928a8a2 216static unsigned long trace_buf_size = TRACE_BUF_SIZE_DEFAULT;
bc0c38d1 217
4fcdae83 218/* trace_types holds a link list of available tracers. */
bc0c38d1 219static struct tracer *trace_types __read_mostly;
4fcdae83
SR
220
221/* current_trace points to the tracer that is currently active */
bc0c38d1 222static struct tracer *current_trace __read_mostly;
4fcdae83
SR
223
224/*
225 * max_tracer_type_len is used to simplify the allocating of
226 * buffers to read userspace tracer names. We keep track of
227 * the longest tracer name registered.
228 */
bc0c38d1
SR
229static int max_tracer_type_len;
230
4fcdae83
SR
231/*
232 * trace_types_lock is used to protect the trace_types list.
233 * This lock is also used to keep user access serialized.
234 * Accesses from userspace will grab this lock while userspace
235 * activities happen inside the kernel.
236 */
bc0c38d1 237static DEFINE_MUTEX(trace_types_lock);
4fcdae83
SR
238
239/* trace_wait is a waitqueue for tasks blocked on trace_poll */
4e655519
IM
240static DECLARE_WAIT_QUEUE_HEAD(trace_wait);
241
ee6bce52 242/* trace_flags holds trace_options default values */
12ef7d44 243unsigned long trace_flags = TRACE_ITER_PRINT_PARENT | TRACE_ITER_PRINTK |
c4a8e8be 244 TRACE_ITER_ANNOTATE | TRACE_ITER_CONTEXT_INFO;
4e655519 245
4fcdae83
SR
246/**
247 * trace_wake_up - wake up tasks waiting for trace input
248 *
249 * Simply wakes up any task that is blocked on the trace_wait
250 * queue. These is used with trace_poll for tasks polling the trace.
251 */
4e655519
IM
252void trace_wake_up(void)
253{
017730c1
IM
254 /*
255 * The runqueue_is_locked() can fail, but this is the best we
256 * have for now:
257 */
258 if (!(trace_flags & TRACE_ITER_BLOCK) && !runqueue_is_locked())
4e655519
IM
259 wake_up(&trace_wait);
260}
bc0c38d1 261
3928a8a2 262static int __init set_buf_size(char *str)
bc0c38d1 263{
3928a8a2 264 unsigned long buf_size;
c6caeeb1
SR
265 int ret;
266
bc0c38d1
SR
267 if (!str)
268 return 0;
3928a8a2 269 ret = strict_strtoul(str, 0, &buf_size);
c6caeeb1 270 /* nr_entries can not be zero */
3928a8a2 271 if (ret < 0 || buf_size == 0)
c6caeeb1 272 return 0;
3928a8a2 273 trace_buf_size = buf_size;
bc0c38d1
SR
274 return 1;
275}
3928a8a2 276__setup("trace_buf_size=", set_buf_size);
bc0c38d1 277
57f50be1
SR
278unsigned long nsecs_to_usecs(unsigned long nsecs)
279{
280 return nsecs / 1000;
281}
282
4fcdae83 283/* These must match the bit postions in trace_iterator_flags */
bc0c38d1
SR
284static const char *trace_options[] = {
285 "print-parent",
286 "sym-offset",
287 "sym-addr",
288 "verbose",
f9896bf3 289 "raw",
5e3ca0ec 290 "hex",
cb0f12aa 291 "bin",
2a2cc8f7 292 "block",
86387f7e 293 "stacktrace",
4ac3ba41 294 "sched-tree",
f09ce573 295 "ftrace_printk",
b2a866f9 296 "ftrace_preempt",
9f029e83 297 "branch",
12ef7d44 298 "annotate",
02b67518 299 "userstacktrace",
b54d3de9 300 "sym-userobj",
66896a85 301 "printk-msg-only",
c4a8e8be 302 "context-info",
bc0c38d1
SR
303 NULL
304};
305
4fcdae83
SR
306/*
307 * ftrace_max_lock is used to protect the swapping of buffers
308 * when taking a max snapshot. The buffers themselves are
309 * protected by per_cpu spinlocks. But the action of the swap
310 * needs its own lock.
311 *
312 * This is defined as a raw_spinlock_t in order to help
313 * with performance when lockdep debugging is enabled.
314 */
92205c23
SR
315static raw_spinlock_t ftrace_max_lock =
316 (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED;
bc0c38d1
SR
317
318/*
319 * Copy the new maximum trace into the separate maximum-trace
320 * structure. (this way the maximum trace is permanently saved,
321 * for later retrieval via /debugfs/tracing/latency_trace)
322 */
e309b41d 323static void
bc0c38d1
SR
324__update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu)
325{
326 struct trace_array_cpu *data = tr->data[cpu];
327
328 max_tr.cpu = cpu;
329 max_tr.time_start = data->preempt_timestamp;
330
331 data = max_tr.data[cpu];
332 data->saved_latency = tracing_max_latency;
333
334 memcpy(data->comm, tsk->comm, TASK_COMM_LEN);
335 data->pid = tsk->pid;
b6dff3ec 336 data->uid = task_uid(tsk);
bc0c38d1
SR
337 data->nice = tsk->static_prio - 20 - MAX_RT_PRIO;
338 data->policy = tsk->policy;
339 data->rt_priority = tsk->rt_priority;
340
341 /* record this tasks comm */
af513098 342 tracing_record_cmdline(tsk);
bc0c38d1
SR
343}
344
6c6c2796
PP
345ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf, size_t cnt)
346{
347 int len;
348 int ret;
349
350 if (s->len <= s->readpos)
351 return -EBUSY;
352
353 len = s->len - s->readpos;
354 if (cnt > len)
355 cnt = len;
356 ret = copy_to_user(ubuf, s->buffer + s->readpos, cnt);
357 if (ret)
358 return -EFAULT;
359
360 s->readpos += len;
361 return cnt;
214023c3
SR
362}
363
3c56819b
EGM
364ssize_t trace_seq_to_buffer(struct trace_seq *s, void *buf, size_t cnt)
365{
366 int len;
367 void *ret;
368
369 if (s->len <= s->readpos)
370 return -EBUSY;
371
372 len = s->len - s->readpos;
373 if (cnt > len)
374 cnt = len;
375 ret = memcpy(buf, s->buffer + s->readpos, cnt);
376 if (!ret)
377 return -EFAULT;
378
379 s->readpos += len;
380 return cnt;
381}
382
e309b41d 383static void
214023c3
SR
384trace_print_seq(struct seq_file *m, struct trace_seq *s)
385{
386 int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len;
387
388 s->buffer[len] = 0;
389 seq_puts(m, s->buffer);
390
f9520750 391 trace_seq_init(s);
214023c3
SR
392}
393
4fcdae83
SR
394/**
395 * update_max_tr - snapshot all trace buffers from global_trace to max_tr
396 * @tr: tracer
397 * @tsk: the task with the latency
398 * @cpu: The cpu that initiated the trace.
399 *
400 * Flip the buffers between the @tr and the max_tr and record information
401 * about which task was the cause of this latency.
402 */
e309b41d 403void
bc0c38d1
SR
404update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu)
405{
3928a8a2 406 struct ring_buffer *buf = tr->buffer;
bc0c38d1 407
4c11d7ae 408 WARN_ON_ONCE(!irqs_disabled());
92205c23 409 __raw_spin_lock(&ftrace_max_lock);
3928a8a2
SR
410
411 tr->buffer = max_tr.buffer;
412 max_tr.buffer = buf;
413
d769041f 414 ftrace_disable_cpu();
3928a8a2 415 ring_buffer_reset(tr->buffer);
d769041f 416 ftrace_enable_cpu();
bc0c38d1
SR
417
418 __update_max_tr(tr, tsk, cpu);
92205c23 419 __raw_spin_unlock(&ftrace_max_lock);
bc0c38d1
SR
420}
421
422/**
423 * update_max_tr_single - only copy one trace over, and reset the rest
424 * @tr - tracer
425 * @tsk - task with the latency
426 * @cpu - the cpu of the buffer to copy.
4fcdae83
SR
427 *
428 * Flip the trace of a single CPU buffer between the @tr and the max_tr.
bc0c38d1 429 */
e309b41d 430void
bc0c38d1
SR
431update_max_tr_single(struct trace_array *tr, struct task_struct *tsk, int cpu)
432{
3928a8a2 433 int ret;
bc0c38d1 434
4c11d7ae 435 WARN_ON_ONCE(!irqs_disabled());
92205c23 436 __raw_spin_lock(&ftrace_max_lock);
bc0c38d1 437
d769041f
SR
438 ftrace_disable_cpu();
439
3928a8a2
SR
440 ring_buffer_reset(max_tr.buffer);
441 ret = ring_buffer_swap_cpu(max_tr.buffer, tr->buffer, cpu);
442
d769041f
SR
443 ftrace_enable_cpu();
444
97b17efe 445 WARN_ON_ONCE(ret && ret != -EAGAIN);
bc0c38d1
SR
446
447 __update_max_tr(tr, tsk, cpu);
92205c23 448 __raw_spin_unlock(&ftrace_max_lock);
bc0c38d1
SR
449}
450
4fcdae83
SR
451/**
452 * register_tracer - register a tracer with the ftrace system.
453 * @type - the plugin for the tracer
454 *
455 * Register a new plugin tracer.
456 */
bc0c38d1 457int register_tracer(struct tracer *type)
e7669b8e
HE
458__releases(kernel_lock)
459__acquires(kernel_lock)
bc0c38d1
SR
460{
461 struct tracer *t;
462 int len;
463 int ret = 0;
464
465 if (!type->name) {
466 pr_info("Tracer must have a name\n");
467 return -1;
468 }
469
86fa2f60
IM
470 /*
471 * When this gets called we hold the BKL which means that
472 * preemption is disabled. Various trace selftests however
473 * need to disable and enable preemption for successful tests.
474 * So we drop the BKL here and grab it after the tests again.
475 */
476 unlock_kernel();
bc0c38d1 477 mutex_lock(&trace_types_lock);
86fa2f60 478
8e1b82e0
FW
479 tracing_selftest_running = true;
480
bc0c38d1
SR
481 for (t = trace_types; t; t = t->next) {
482 if (strcmp(type->name, t->name) == 0) {
483 /* already found */
484 pr_info("Trace %s already registered\n",
485 type->name);
486 ret = -1;
487 goto out;
488 }
489 }
490
adf9f195
FW
491 if (!type->set_flag)
492 type->set_flag = &dummy_set_flag;
493 if (!type->flags)
494 type->flags = &dummy_tracer_flags;
495 else
496 if (!type->flags->opts)
497 type->flags->opts = dummy_tracer_opt;
6eaaa5d5
FW
498 if (!type->wait_pipe)
499 type->wait_pipe = default_wait_pipe;
500
adf9f195 501
60a11774 502#ifdef CONFIG_FTRACE_STARTUP_TEST
b2821ae6 503 if (type->selftest && !tracing_selftest_disabled) {
60a11774 504 struct tracer *saved_tracer = current_trace;
60a11774 505 struct trace_array *tr = &global_trace;
60a11774 506 int i;
ff32504f 507
60a11774
SR
508 /*
509 * Run a selftest on this tracer.
510 * Here we reset the trace buffer, and set the current
511 * tracer to be this tracer. The tracer can then run some
512 * internal tracing to verify that everything is in order.
513 * If we fail, we do not register this tracer.
514 */
86fa2f60 515 for_each_tracing_cpu(i)
3928a8a2 516 tracing_reset(tr, i);
86fa2f60 517
60a11774 518 current_trace = type;
60a11774
SR
519 /* the test is responsible for initializing and enabling */
520 pr_info("Testing tracer %s: ", type->name);
521 ret = type->selftest(type, tr);
522 /* the test is responsible for resetting too */
523 current_trace = saved_tracer;
60a11774
SR
524 if (ret) {
525 printk(KERN_CONT "FAILED!\n");
526 goto out;
527 }
1d4db00a 528 /* Only reset on passing, to avoid touching corrupted buffers */
86fa2f60 529 for_each_tracing_cpu(i)
3928a8a2 530 tracing_reset(tr, i);
86fa2f60 531
60a11774
SR
532 printk(KERN_CONT "PASSED\n");
533 }
534#endif
535
bc0c38d1
SR
536 type->next = trace_types;
537 trace_types = type;
538 len = strlen(type->name);
539 if (len > max_tracer_type_len)
540 max_tracer_type_len = len;
60a11774 541
bc0c38d1 542 out:
8e1b82e0 543 tracing_selftest_running = false;
bc0c38d1
SR
544 mutex_unlock(&trace_types_lock);
545
dac74940
SR
546 if (ret || !default_bootup_tracer)
547 goto out_unlock;
548
549 if (strncmp(default_bootup_tracer, type->name, BOOTUP_TRACER_SIZE))
550 goto out_unlock;
551
552 printk(KERN_INFO "Starting tracer '%s'\n", type->name);
553 /* Do we want this tracer to start on bootup? */
554 tracing_set_tracer(type->name);
555 default_bootup_tracer = NULL;
556 /* disable other selftests, since this will break it. */
557 tracing_selftest_disabled = 1;
b2821ae6 558#ifdef CONFIG_FTRACE_STARTUP_TEST
dac74940
SR
559 printk(KERN_INFO "Disabling FTRACE selftests due to running tracer '%s'\n",
560 type->name);
b2821ae6 561#endif
b2821ae6 562
dac74940 563 out_unlock:
b2821ae6 564 lock_kernel();
bc0c38d1
SR
565 return ret;
566}
567
568void unregister_tracer(struct tracer *type)
569{
570 struct tracer **t;
571 int len;
572
573 mutex_lock(&trace_types_lock);
574 for (t = &trace_types; *t; t = &(*t)->next) {
575 if (*t == type)
576 goto found;
577 }
578 pr_info("Trace %s not registered\n", type->name);
579 goto out;
580
581 found:
582 *t = (*t)->next;
b5db03c4
ACM
583
584 if (type == current_trace && tracer_enabled) {
585 tracer_enabled = 0;
586 tracing_stop();
587 if (current_trace->stop)
588 current_trace->stop(&global_trace);
589 current_trace = &nop_trace;
590 }
591
bc0c38d1
SR
592 if (strlen(type->name) != max_tracer_type_len)
593 goto out;
594
595 max_tracer_type_len = 0;
596 for (t = &trace_types; *t; t = &(*t)->next) {
597 len = strlen((*t)->name);
598 if (len > max_tracer_type_len)
599 max_tracer_type_len = len;
600 }
601 out:
602 mutex_unlock(&trace_types_lock);
603}
604
3928a8a2 605void tracing_reset(struct trace_array *tr, int cpu)
bc0c38d1 606{
d769041f 607 ftrace_disable_cpu();
3928a8a2 608 ring_buffer_reset_cpu(tr->buffer, cpu);
d769041f 609 ftrace_enable_cpu();
bc0c38d1
SR
610}
611
213cc060
PE
612void tracing_reset_online_cpus(struct trace_array *tr)
613{
614 int cpu;
615
616 tr->time_start = ftrace_now(tr->cpu);
617
618 for_each_online_cpu(cpu)
619 tracing_reset(tr, cpu);
620}
621
bc0c38d1
SR
622#define SAVED_CMDLINES 128
623static unsigned map_pid_to_cmdline[PID_MAX_DEFAULT+1];
624static unsigned map_cmdline_to_pid[SAVED_CMDLINES];
625static char saved_cmdlines[SAVED_CMDLINES][TASK_COMM_LEN];
626static int cmdline_idx;
627static DEFINE_SPINLOCK(trace_cmdline_lock);
25b0b44a 628
25b0b44a 629/* temporary disable recording */
4fd27358 630static atomic_t trace_record_cmdline_disabled __read_mostly;
bc0c38d1
SR
631
632static void trace_init_cmdlines(void)
633{
634 memset(&map_pid_to_cmdline, -1, sizeof(map_pid_to_cmdline));
635 memset(&map_cmdline_to_pid, -1, sizeof(map_cmdline_to_pid));
636 cmdline_idx = 0;
637}
638
0f048701
SR
639static int trace_stop_count;
640static DEFINE_SPINLOCK(tracing_start_lock);
641
69bb54ec
SR
642/**
643 * ftrace_off_permanent - disable all ftrace code permanently
644 *
645 * This should only be called when a serious anomally has
646 * been detected. This will turn off the function tracing,
647 * ring buffers, and other tracing utilites. It takes no
648 * locks and can be called from any context.
649 */
650void ftrace_off_permanent(void)
651{
652 tracing_disabled = 1;
653 ftrace_stop();
654 tracing_off_permanent();
655}
656
0f048701
SR
657/**
658 * tracing_start - quick start of the tracer
659 *
660 * If tracing is enabled but was stopped by tracing_stop,
661 * this will start the tracer back up.
662 */
663void tracing_start(void)
664{
665 struct ring_buffer *buffer;
666 unsigned long flags;
667
668 if (tracing_disabled)
669 return;
670
671 spin_lock_irqsave(&tracing_start_lock, flags);
b06a8301
SR
672 if (--trace_stop_count) {
673 if (trace_stop_count < 0) {
674 /* Someone screwed up their debugging */
675 WARN_ON_ONCE(1);
676 trace_stop_count = 0;
677 }
0f048701
SR
678 goto out;
679 }
680
681
682 buffer = global_trace.buffer;
683 if (buffer)
684 ring_buffer_record_enable(buffer);
685
686 buffer = max_tr.buffer;
687 if (buffer)
688 ring_buffer_record_enable(buffer);
689
690 ftrace_start();
691 out:
692 spin_unlock_irqrestore(&tracing_start_lock, flags);
693}
694
695/**
696 * tracing_stop - quick stop of the tracer
697 *
698 * Light weight way to stop tracing. Use in conjunction with
699 * tracing_start.
700 */
701void tracing_stop(void)
702{
703 struct ring_buffer *buffer;
704 unsigned long flags;
705
706 ftrace_stop();
707 spin_lock_irqsave(&tracing_start_lock, flags);
708 if (trace_stop_count++)
709 goto out;
710
711 buffer = global_trace.buffer;
712 if (buffer)
713 ring_buffer_record_disable(buffer);
714
715 buffer = max_tr.buffer;
716 if (buffer)
717 ring_buffer_record_disable(buffer);
718
719 out:
720 spin_unlock_irqrestore(&tracing_start_lock, flags);
721}
722
e309b41d 723void trace_stop_cmdline_recording(void);
bc0c38d1 724
e309b41d 725static void trace_save_cmdline(struct task_struct *tsk)
bc0c38d1
SR
726{
727 unsigned map;
728 unsigned idx;
729
730 if (!tsk->pid || unlikely(tsk->pid > PID_MAX_DEFAULT))
731 return;
732
733 /*
734 * It's not the end of the world if we don't get
735 * the lock, but we also don't want to spin
736 * nor do we want to disable interrupts,
737 * so if we miss here, then better luck next time.
738 */
739 if (!spin_trylock(&trace_cmdline_lock))
740 return;
741
742 idx = map_pid_to_cmdline[tsk->pid];
743 if (idx >= SAVED_CMDLINES) {
744 idx = (cmdline_idx + 1) % SAVED_CMDLINES;
745
746 map = map_cmdline_to_pid[idx];
747 if (map <= PID_MAX_DEFAULT)
748 map_pid_to_cmdline[map] = (unsigned)-1;
749
750 map_pid_to_cmdline[tsk->pid] = idx;
751
752 cmdline_idx = idx;
753 }
754
755 memcpy(&saved_cmdlines[idx], tsk->comm, TASK_COMM_LEN);
756
757 spin_unlock(&trace_cmdline_lock);
758}
759
660c7f9b 760char *trace_find_cmdline(int pid)
bc0c38d1
SR
761{
762 char *cmdline = "<...>";
763 unsigned map;
764
765 if (!pid)
766 return "<idle>";
767
768 if (pid > PID_MAX_DEFAULT)
769 goto out;
770
771 map = map_pid_to_cmdline[pid];
772 if (map >= SAVED_CMDLINES)
773 goto out;
774
775 cmdline = saved_cmdlines[map];
776
777 out:
778 return cmdline;
779}
780
e309b41d 781void tracing_record_cmdline(struct task_struct *tsk)
bc0c38d1
SR
782{
783 if (atomic_read(&trace_record_cmdline_disabled))
784 return;
785
786 trace_save_cmdline(tsk);
787}
788
45dcd8b8 789void
38697053
SR
790tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags,
791 int pc)
bc0c38d1
SR
792{
793 struct task_struct *tsk = current;
bc0c38d1 794
777e208d
SR
795 entry->preempt_count = pc & 0xff;
796 entry->pid = (tsk) ? tsk->pid : 0;
b54d3de9 797 entry->tgid = (tsk) ? tsk->tgid : 0;
777e208d 798 entry->flags =
9244489a 799#ifdef CONFIG_TRACE_IRQFLAGS_SUPPORT
2e2ca155 800 (irqs_disabled_flags(flags) ? TRACE_FLAG_IRQS_OFF : 0) |
9244489a
SR
801#else
802 TRACE_FLAG_IRQS_NOSUPPORT |
803#endif
bc0c38d1
SR
804 ((pc & HARDIRQ_MASK) ? TRACE_FLAG_HARDIRQ : 0) |
805 ((pc & SOFTIRQ_MASK) ? TRACE_FLAG_SOFTIRQ : 0) |
806 (need_resched() ? TRACE_FLAG_NEED_RESCHED : 0);
807}
808
51a763dd
ACM
809struct ring_buffer_event *trace_buffer_lock_reserve(struct trace_array *tr,
810 unsigned char type,
811 unsigned long len,
812 unsigned long flags, int pc)
813{
814 struct ring_buffer_event *event;
815
816 event = ring_buffer_lock_reserve(tr->buffer, len);
817 if (event != NULL) {
818 struct trace_entry *ent = ring_buffer_event_data(event);
819
820 tracing_generic_entry_update(ent, flags, pc);
821 ent->type = type;
822 }
823
824 return event;
825}
826static void ftrace_trace_stack(struct trace_array *tr,
827 unsigned long flags, int skip, int pc);
828static void ftrace_trace_userstack(struct trace_array *tr,
829 unsigned long flags, int pc);
830
831void trace_buffer_unlock_commit(struct trace_array *tr,
832 struct ring_buffer_event *event,
833 unsigned long flags, int pc)
834{
835 ring_buffer_unlock_commit(tr->buffer, event);
836
837 ftrace_trace_stack(tr, flags, 6, pc);
838 ftrace_trace_userstack(tr, flags, pc);
839 trace_wake_up();
840}
841
ef5580d0
SR
842struct ring_buffer_event *
843trace_current_buffer_lock_reserve(unsigned char type, unsigned long len,
844 unsigned long flags, int pc)
845{
846 return trace_buffer_lock_reserve(&global_trace,
847 type, len, flags, pc);
848}
849
850void trace_current_buffer_unlock_commit(struct ring_buffer_event *event,
851 unsigned long flags, int pc)
852{
853 return trace_buffer_unlock_commit(&global_trace, event, flags, pc);
854}
855
e309b41d 856void
7be42151 857trace_function(struct trace_array *tr,
38697053
SR
858 unsigned long ip, unsigned long parent_ip, unsigned long flags,
859 int pc)
bc0c38d1 860{
3928a8a2 861 struct ring_buffer_event *event;
777e208d 862 struct ftrace_entry *entry;
bc0c38d1 863
d769041f
SR
864 /* If we are reading the ring buffer, don't trace */
865 if (unlikely(local_read(&__get_cpu_var(ftrace_cpu_disabled))))
866 return;
867
51a763dd
ACM
868 event = trace_buffer_lock_reserve(tr, TRACE_FN, sizeof(*entry),
869 flags, pc);
3928a8a2
SR
870 if (!event)
871 return;
872 entry = ring_buffer_event_data(event);
777e208d
SR
873 entry->ip = ip;
874 entry->parent_ip = parent_ip;
0a987751 875 ring_buffer_unlock_commit(tr->buffer, event);
bc0c38d1
SR
876}
877
fb52607a 878#ifdef CONFIG_FUNCTION_GRAPH_TRACER
287b6e68 879static void __trace_graph_entry(struct trace_array *tr,
287b6e68
FW
880 struct ftrace_graph_ent *trace,
881 unsigned long flags,
882 int pc)
883{
884 struct ring_buffer_event *event;
885 struct ftrace_graph_ent_entry *entry;
287b6e68
FW
886
887 if (unlikely(local_read(&__get_cpu_var(ftrace_cpu_disabled))))
888 return;
889
51a763dd
ACM
890 event = trace_buffer_lock_reserve(&global_trace, TRACE_GRAPH_ENT,
891 sizeof(*entry), flags, pc);
287b6e68
FW
892 if (!event)
893 return;
894 entry = ring_buffer_event_data(event);
287b6e68 895 entry->graph_ent = *trace;
0a987751 896 ring_buffer_unlock_commit(global_trace.buffer, event);
287b6e68
FW
897}
898
899static void __trace_graph_return(struct trace_array *tr,
fb52607a 900 struct ftrace_graph_ret *trace,
15e6cb36
FW
901 unsigned long flags,
902 int pc)
903{
904 struct ring_buffer_event *event;
287b6e68 905 struct ftrace_graph_ret_entry *entry;
15e6cb36
FW
906
907 if (unlikely(local_read(&__get_cpu_var(ftrace_cpu_disabled))))
908 return;
909
51a763dd
ACM
910 event = trace_buffer_lock_reserve(&global_trace, TRACE_GRAPH_RET,
911 sizeof(*entry), flags, pc);
15e6cb36
FW
912 if (!event)
913 return;
914 entry = ring_buffer_event_data(event);
287b6e68 915 entry->ret = *trace;
0a987751 916 ring_buffer_unlock_commit(global_trace.buffer, event);
15e6cb36
FW
917}
918#endif
919
e309b41d 920void
2e0f5761 921ftrace(struct trace_array *tr, struct trace_array_cpu *data,
38697053
SR
922 unsigned long ip, unsigned long parent_ip, unsigned long flags,
923 int pc)
2e0f5761
IM
924{
925 if (likely(!atomic_read(&data->disabled)))
7be42151 926 trace_function(tr, ip, parent_ip, flags, pc);
2e0f5761
IM
927}
928
53614991 929static void __ftrace_trace_stack(struct trace_array *tr,
53614991
SR
930 unsigned long flags,
931 int skip, int pc)
86387f7e 932{
c2c80529 933#ifdef CONFIG_STACKTRACE
3928a8a2 934 struct ring_buffer_event *event;
777e208d 935 struct stack_entry *entry;
86387f7e
IM
936 struct stack_trace trace;
937
51a763dd
ACM
938 event = trace_buffer_lock_reserve(tr, TRACE_STACK,
939 sizeof(*entry), flags, pc);
3928a8a2
SR
940 if (!event)
941 return;
942 entry = ring_buffer_event_data(event);
777e208d 943 memset(&entry->caller, 0, sizeof(entry->caller));
86387f7e
IM
944
945 trace.nr_entries = 0;
946 trace.max_entries = FTRACE_STACK_ENTRIES;
947 trace.skip = skip;
777e208d 948 trace.entries = entry->caller;
86387f7e
IM
949
950 save_stack_trace(&trace);
0a987751 951 ring_buffer_unlock_commit(tr->buffer, event);
c2c80529 952#endif
f0a920d5
IM
953}
954
53614991 955static void ftrace_trace_stack(struct trace_array *tr,
53614991
SR
956 unsigned long flags,
957 int skip, int pc)
958{
959 if (!(trace_flags & TRACE_ITER_STACKTRACE))
960 return;
961
7be42151 962 __ftrace_trace_stack(tr, flags, skip, pc);
53614991
SR
963}
964
38697053 965void __trace_stack(struct trace_array *tr,
38697053 966 unsigned long flags,
53614991 967 int skip, int pc)
38697053 968{
7be42151 969 __ftrace_trace_stack(tr, flags, skip, pc);
38697053
SR
970}
971
02b67518 972static void ftrace_trace_userstack(struct trace_array *tr,
7be42151 973 unsigned long flags, int pc)
02b67518 974{
c7425acb 975#ifdef CONFIG_STACKTRACE
8d7c6a96 976 struct ring_buffer_event *event;
02b67518
TE
977 struct userstack_entry *entry;
978 struct stack_trace trace;
02b67518
TE
979
980 if (!(trace_flags & TRACE_ITER_USERSTACKTRACE))
981 return;
982
51a763dd
ACM
983 event = trace_buffer_lock_reserve(tr, TRACE_USER_STACK,
984 sizeof(*entry), flags, pc);
02b67518
TE
985 if (!event)
986 return;
987 entry = ring_buffer_event_data(event);
02b67518
TE
988
989 memset(&entry->caller, 0, sizeof(entry->caller));
990
991 trace.nr_entries = 0;
992 trace.max_entries = FTRACE_STACK_ENTRIES;
993 trace.skip = 0;
994 trace.entries = entry->caller;
995
996 save_stack_trace_user(&trace);
0a987751 997 ring_buffer_unlock_commit(tr->buffer, event);
c7425acb 998#endif
02b67518
TE
999}
1000
4fd27358
HE
1001#ifdef UNUSED
1002static void __trace_userstack(struct trace_array *tr, unsigned long flags)
02b67518 1003{
7be42151 1004 ftrace_trace_userstack(tr, flags, preempt_count());
02b67518 1005}
4fd27358 1006#endif /* UNUSED */
02b67518 1007
38697053 1008static void
7be42151 1009ftrace_trace_special(void *__tr,
38697053
SR
1010 unsigned long arg1, unsigned long arg2, unsigned long arg3,
1011 int pc)
a4feb834 1012{
3928a8a2 1013 struct ring_buffer_event *event;
a4feb834 1014 struct trace_array *tr = __tr;
777e208d 1015 struct special_entry *entry;
a4feb834 1016
51a763dd
ACM
1017 event = trace_buffer_lock_reserve(tr, TRACE_SPECIAL,
1018 sizeof(*entry), 0, pc);
3928a8a2
SR
1019 if (!event)
1020 return;
1021 entry = ring_buffer_event_data(event);
777e208d
SR
1022 entry->arg1 = arg1;
1023 entry->arg2 = arg2;
1024 entry->arg3 = arg3;
51a763dd 1025 trace_buffer_unlock_commit(tr, event, 0, pc);
a4feb834
IM
1026}
1027
38697053
SR
1028void
1029__trace_special(void *__tr, void *__data,
1030 unsigned long arg1, unsigned long arg2, unsigned long arg3)
1031{
7be42151 1032 ftrace_trace_special(__tr, arg1, arg2, arg3, preempt_count());
38697053
SR
1033}
1034
e309b41d 1035void
bc0c38d1 1036tracing_sched_switch_trace(struct trace_array *tr,
86387f7e
IM
1037 struct task_struct *prev,
1038 struct task_struct *next,
38697053 1039 unsigned long flags, int pc)
bc0c38d1 1040{
3928a8a2 1041 struct ring_buffer_event *event;
777e208d 1042 struct ctx_switch_entry *entry;
bc0c38d1 1043
51a763dd
ACM
1044 event = trace_buffer_lock_reserve(tr, TRACE_CTX,
1045 sizeof(*entry), flags, pc);
3928a8a2
SR
1046 if (!event)
1047 return;
1048 entry = ring_buffer_event_data(event);
777e208d
SR
1049 entry->prev_pid = prev->pid;
1050 entry->prev_prio = prev->prio;
1051 entry->prev_state = prev->state;
1052 entry->next_pid = next->pid;
1053 entry->next_prio = next->prio;
1054 entry->next_state = next->state;
1055 entry->next_cpu = task_cpu(next);
51a763dd 1056 trace_buffer_unlock_commit(tr, event, flags, pc);
bc0c38d1
SR
1057}
1058
57422797
IM
1059void
1060tracing_sched_wakeup_trace(struct trace_array *tr,
86387f7e
IM
1061 struct task_struct *wakee,
1062 struct task_struct *curr,
38697053 1063 unsigned long flags, int pc)
57422797 1064{
3928a8a2 1065 struct ring_buffer_event *event;
777e208d 1066 struct ctx_switch_entry *entry;
57422797 1067
51a763dd
ACM
1068 event = trace_buffer_lock_reserve(tr, TRACE_WAKE,
1069 sizeof(*entry), flags, pc);
3928a8a2
SR
1070 if (!event)
1071 return;
1072 entry = ring_buffer_event_data(event);
777e208d
SR
1073 entry->prev_pid = curr->pid;
1074 entry->prev_prio = curr->prio;
1075 entry->prev_state = curr->state;
1076 entry->next_pid = wakee->pid;
1077 entry->next_prio = wakee->prio;
1078 entry->next_state = wakee->state;
1079 entry->next_cpu = task_cpu(wakee);
6eaaa5d5
FW
1080
1081 ring_buffer_unlock_commit(tr->buffer, event);
1082 ftrace_trace_stack(tr, flags, 6, pc);
1083 ftrace_trace_userstack(tr, flags, pc);
57422797
IM
1084}
1085
4902f884
SR
1086void
1087ftrace_special(unsigned long arg1, unsigned long arg2, unsigned long arg3)
1088{
1089 struct trace_array *tr = &global_trace;
1090 struct trace_array_cpu *data;
5aa1ba6a 1091 unsigned long flags;
4902f884 1092 int cpu;
38697053 1093 int pc;
4902f884 1094
c76f0694 1095 if (tracing_disabled)
4902f884
SR
1096 return;
1097
38697053 1098 pc = preempt_count();
5aa1ba6a 1099 local_irq_save(flags);
4902f884
SR
1100 cpu = raw_smp_processor_id();
1101 data = tr->data[cpu];
4902f884 1102
5aa1ba6a 1103 if (likely(atomic_inc_return(&data->disabled) == 1))
7be42151 1104 ftrace_trace_special(tr, arg1, arg2, arg3, pc);
4902f884 1105
5aa1ba6a
SR
1106 atomic_dec(&data->disabled);
1107 local_irq_restore(flags);
4902f884
SR
1108}
1109
fb52607a 1110#ifdef CONFIG_FUNCTION_GRAPH_TRACER
e49dc19c 1111int trace_graph_entry(struct ftrace_graph_ent *trace)
15e6cb36
FW
1112{
1113 struct trace_array *tr = &global_trace;
1114 struct trace_array_cpu *data;
1115 unsigned long flags;
1116 long disabled;
1117 int cpu;
1118 int pc;
1119
804a6851
SR
1120 if (!ftrace_trace_task(current))
1121 return 0;
1122
ea4e2bc4
SR
1123 if (!ftrace_graph_addr(trace->func))
1124 return 0;
1125
a5e25883 1126 local_irq_save(flags);
15e6cb36
FW
1127 cpu = raw_smp_processor_id();
1128 data = tr->data[cpu];
1129 disabled = atomic_inc_return(&data->disabled);
1130 if (likely(disabled == 1)) {
1131 pc = preempt_count();
7be42151 1132 __trace_graph_entry(tr, trace, flags, pc);
287b6e68 1133 }
ea4e2bc4
SR
1134 /* Only do the atomic if it is not already set */
1135 if (!test_tsk_trace_graph(current))
1136 set_tsk_trace_graph(current);
287b6e68 1137 atomic_dec(&data->disabled);
a5e25883 1138 local_irq_restore(flags);
e49dc19c
SR
1139
1140 return 1;
287b6e68
FW
1141}
1142
1143void trace_graph_return(struct ftrace_graph_ret *trace)
1144{
1145 struct trace_array *tr = &global_trace;
1146 struct trace_array_cpu *data;
1147 unsigned long flags;
1148 long disabled;
1149 int cpu;
1150 int pc;
1151
a5e25883 1152 local_irq_save(flags);
287b6e68
FW
1153 cpu = raw_smp_processor_id();
1154 data = tr->data[cpu];
1155 disabled = atomic_inc_return(&data->disabled);
1156 if (likely(disabled == 1)) {
1157 pc = preempt_count();
7be42151 1158 __trace_graph_return(tr, trace, flags, pc);
15e6cb36 1159 }
ea4e2bc4
SR
1160 if (!trace->depth)
1161 clear_tsk_trace_graph(current);
15e6cb36 1162 atomic_dec(&data->disabled);
a5e25883 1163 local_irq_restore(flags);
15e6cb36 1164}
fb52607a 1165#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
15e6cb36 1166
bc0c38d1
SR
1167enum trace_file_type {
1168 TRACE_FILE_LAT_FMT = 1,
12ef7d44 1169 TRACE_FILE_ANNOTATE = 2,
bc0c38d1
SR
1170};
1171
e2ac8ef5 1172static void trace_iterator_increment(struct trace_iterator *iter)
5a90f577 1173{
d769041f
SR
1174 /* Don't allow ftrace to trace into the ring buffers */
1175 ftrace_disable_cpu();
1176
5a90f577 1177 iter->idx++;
d769041f
SR
1178 if (iter->buffer_iter[iter->cpu])
1179 ring_buffer_read(iter->buffer_iter[iter->cpu], NULL);
1180
1181 ftrace_enable_cpu();
5a90f577
SR
1182}
1183
e309b41d 1184static struct trace_entry *
3928a8a2 1185peek_next_entry(struct trace_iterator *iter, int cpu, u64 *ts)
dd0e545f 1186{
3928a8a2
SR
1187 struct ring_buffer_event *event;
1188 struct ring_buffer_iter *buf_iter = iter->buffer_iter[cpu];
dd0e545f 1189
d769041f
SR
1190 /* Don't allow ftrace to trace into the ring buffers */
1191 ftrace_disable_cpu();
1192
1193 if (buf_iter)
1194 event = ring_buffer_iter_peek(buf_iter, ts);
1195 else
1196 event = ring_buffer_peek(iter->tr->buffer, cpu, ts);
1197
1198 ftrace_enable_cpu();
1199
3928a8a2 1200 return event ? ring_buffer_event_data(event) : NULL;
dd0e545f 1201}
d769041f 1202
dd0e545f 1203static struct trace_entry *
3928a8a2 1204__find_next_entry(struct trace_iterator *iter, int *ent_cpu, u64 *ent_ts)
bc0c38d1 1205{
3928a8a2 1206 struct ring_buffer *buffer = iter->tr->buffer;
bc0c38d1 1207 struct trace_entry *ent, *next = NULL;
b04cc6b1 1208 int cpu_file = iter->cpu_file;
3928a8a2 1209 u64 next_ts = 0, ts;
bc0c38d1
SR
1210 int next_cpu = -1;
1211 int cpu;
1212
b04cc6b1
FW
1213 /*
1214 * If we are in a per_cpu trace file, don't bother by iterating over
1215 * all cpu and peek directly.
1216 */
1217 if (cpu_file > TRACE_PIPE_ALL_CPU) {
1218 if (ring_buffer_empty_cpu(buffer, cpu_file))
1219 return NULL;
1220 ent = peek_next_entry(iter, cpu_file, ent_ts);
1221 if (ent_cpu)
1222 *ent_cpu = cpu_file;
1223
1224 return ent;
1225 }
1226
ab46428c 1227 for_each_tracing_cpu(cpu) {
dd0e545f 1228
3928a8a2
SR
1229 if (ring_buffer_empty_cpu(buffer, cpu))
1230 continue;
dd0e545f 1231
3928a8a2 1232 ent = peek_next_entry(iter, cpu, &ts);
dd0e545f 1233
cdd31cd2
IM
1234 /*
1235 * Pick the entry with the smallest timestamp:
1236 */
3928a8a2 1237 if (ent && (!next || ts < next_ts)) {
bc0c38d1
SR
1238 next = ent;
1239 next_cpu = cpu;
3928a8a2 1240 next_ts = ts;
bc0c38d1
SR
1241 }
1242 }
1243
1244 if (ent_cpu)
1245 *ent_cpu = next_cpu;
1246
3928a8a2
SR
1247 if (ent_ts)
1248 *ent_ts = next_ts;
1249
bc0c38d1
SR
1250 return next;
1251}
1252
dd0e545f 1253/* Find the next real entry, without updating the iterator itself */
c4a8e8be
FW
1254struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
1255 int *ent_cpu, u64 *ent_ts)
bc0c38d1 1256{
3928a8a2 1257 return __find_next_entry(iter, ent_cpu, ent_ts);
dd0e545f
SR
1258}
1259
1260/* Find the next real entry, and increment the iterator to the next entry */
1261static void *find_next_entry_inc(struct trace_iterator *iter)
1262{
3928a8a2 1263 iter->ent = __find_next_entry(iter, &iter->cpu, &iter->ts);
dd0e545f 1264
3928a8a2 1265 if (iter->ent)
e2ac8ef5 1266 trace_iterator_increment(iter);
dd0e545f 1267
3928a8a2 1268 return iter->ent ? iter : NULL;
b3806b43 1269}
bc0c38d1 1270
e309b41d 1271static void trace_consume(struct trace_iterator *iter)
b3806b43 1272{
d769041f
SR
1273 /* Don't allow ftrace to trace into the ring buffers */
1274 ftrace_disable_cpu();
3928a8a2 1275 ring_buffer_consume(iter->tr->buffer, iter->cpu, &iter->ts);
d769041f 1276 ftrace_enable_cpu();
bc0c38d1
SR
1277}
1278
e309b41d 1279static void *s_next(struct seq_file *m, void *v, loff_t *pos)
bc0c38d1
SR
1280{
1281 struct trace_iterator *iter = m->private;
bc0c38d1 1282 int i = (int)*pos;
4e3c3333 1283 void *ent;
bc0c38d1
SR
1284
1285 (*pos)++;
1286
1287 /* can't go backwards */
1288 if (iter->idx > i)
1289 return NULL;
1290
1291 if (iter->idx < 0)
1292 ent = find_next_entry_inc(iter);
1293 else
1294 ent = iter;
1295
1296 while (ent && iter->idx < i)
1297 ent = find_next_entry_inc(iter);
1298
1299 iter->pos = *pos;
1300
bc0c38d1
SR
1301 return ent;
1302}
1303
d7350c3f
FW
1304/*
1305 * No necessary locking here. The worst thing which can
1306 * happen is loosing events consumed at the same time
1307 * by a trace_pipe reader.
1308 * Other than that, we don't risk to crash the ring buffer
1309 * because it serializes the readers.
1310 *
1311 * The current tracer is copied to avoid a global locking
1312 * all around.
1313 */
bc0c38d1
SR
1314static void *s_start(struct seq_file *m, loff_t *pos)
1315{
1316 struct trace_iterator *iter = m->private;
d7350c3f 1317 static struct tracer *old_tracer;
b04cc6b1 1318 int cpu_file = iter->cpu_file;
bc0c38d1
SR
1319 void *p = NULL;
1320 loff_t l = 0;
3928a8a2 1321 int cpu;
bc0c38d1 1322
d7350c3f 1323 /* copy the tracer to avoid using a global lock all around */
bc0c38d1 1324 mutex_lock(&trace_types_lock);
d7350c3f
FW
1325 if (unlikely(old_tracer != current_trace && current_trace)) {
1326 old_tracer = current_trace;
1327 *iter->trace = *current_trace;
d15f57f2 1328 }
d7350c3f 1329 mutex_unlock(&trace_types_lock);
bc0c38d1
SR
1330
1331 atomic_inc(&trace_record_cmdline_disabled);
1332
bc0c38d1
SR
1333 if (*pos != iter->pos) {
1334 iter->ent = NULL;
1335 iter->cpu = 0;
1336 iter->idx = -1;
1337
d769041f
SR
1338 ftrace_disable_cpu();
1339
b04cc6b1
FW
1340 if (cpu_file == TRACE_PIPE_ALL_CPU) {
1341 for_each_tracing_cpu(cpu)
1342 ring_buffer_iter_reset(iter->buffer_iter[cpu]);
1343 } else
1344 ring_buffer_iter_reset(iter->buffer_iter[cpu_file]);
1345
bc0c38d1 1346
d769041f
SR
1347 ftrace_enable_cpu();
1348
bc0c38d1
SR
1349 for (p = iter; p && l < *pos; p = s_next(m, p, &l))
1350 ;
1351
1352 } else {
4c11d7ae 1353 l = *pos - 1;
bc0c38d1
SR
1354 p = s_next(m, p, &l);
1355 }
1356
1357 return p;
1358}
1359
1360static void s_stop(struct seq_file *m, void *p)
1361{
bc0c38d1 1362 atomic_dec(&trace_record_cmdline_disabled);
bc0c38d1
SR
1363}
1364
e309b41d 1365static void print_lat_help_header(struct seq_file *m)
bc0c38d1 1366{
a6168353
ME
1367 seq_puts(m, "# _------=> CPU# \n");
1368 seq_puts(m, "# / _-----=> irqs-off \n");
1369 seq_puts(m, "# | / _----=> need-resched \n");
1370 seq_puts(m, "# || / _---=> hardirq/softirq \n");
1371 seq_puts(m, "# ||| / _--=> preempt-depth \n");
1372 seq_puts(m, "# |||| / \n");
1373 seq_puts(m, "# ||||| delay \n");
1374 seq_puts(m, "# cmd pid ||||| time | caller \n");
1375 seq_puts(m, "# \\ / ||||| \\ | / \n");
bc0c38d1
SR
1376}
1377
e309b41d 1378static void print_func_help_header(struct seq_file *m)
bc0c38d1 1379{
a6168353
ME
1380 seq_puts(m, "# TASK-PID CPU# TIMESTAMP FUNCTION\n");
1381 seq_puts(m, "# | | | | |\n");
bc0c38d1
SR
1382}
1383
1384
e309b41d 1385static void
bc0c38d1
SR
1386print_trace_header(struct seq_file *m, struct trace_iterator *iter)
1387{
1388 unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
1389 struct trace_array *tr = iter->tr;
1390 struct trace_array_cpu *data = tr->data[tr->cpu];
1391 struct tracer *type = current_trace;
3928a8a2
SR
1392 unsigned long total;
1393 unsigned long entries;
bc0c38d1
SR
1394 const char *name = "preemption";
1395
1396 if (type)
1397 name = type->name;
1398
3928a8a2
SR
1399 entries = ring_buffer_entries(iter->tr->buffer);
1400 total = entries +
1401 ring_buffer_overruns(iter->tr->buffer);
bc0c38d1
SR
1402
1403 seq_printf(m, "%s latency trace v1.1.5 on %s\n",
1404 name, UTS_RELEASE);
1405 seq_puts(m, "-----------------------------------"
1406 "---------------------------------\n");
1407 seq_printf(m, " latency: %lu us, #%lu/%lu, CPU#%d |"
1408 " (M:%s VP:%d, KP:%d, SP:%d HP:%d",
57f50be1 1409 nsecs_to_usecs(data->saved_latency),
bc0c38d1 1410 entries,
4c11d7ae 1411 total,
bc0c38d1
SR
1412 tr->cpu,
1413#if defined(CONFIG_PREEMPT_NONE)
1414 "server",
1415#elif defined(CONFIG_PREEMPT_VOLUNTARY)
1416 "desktop",
b5c21b45 1417#elif defined(CONFIG_PREEMPT)
bc0c38d1
SR
1418 "preempt",
1419#else
1420 "unknown",
1421#endif
1422 /* These are reserved for later use */
1423 0, 0, 0, 0);
1424#ifdef CONFIG_SMP
1425 seq_printf(m, " #P:%d)\n", num_online_cpus());
1426#else
1427 seq_puts(m, ")\n");
1428#endif
1429 seq_puts(m, " -----------------\n");
1430 seq_printf(m, " | task: %.16s-%d "
1431 "(uid:%d nice:%ld policy:%ld rt_prio:%ld)\n",
1432 data->comm, data->pid, data->uid, data->nice,
1433 data->policy, data->rt_priority);
1434 seq_puts(m, " -----------------\n");
1435
1436 if (data->critical_start) {
1437 seq_puts(m, " => started at: ");
214023c3
SR
1438 seq_print_ip_sym(&iter->seq, data->critical_start, sym_flags);
1439 trace_print_seq(m, &iter->seq);
bc0c38d1 1440 seq_puts(m, "\n => ended at: ");
214023c3
SR
1441 seq_print_ip_sym(&iter->seq, data->critical_end, sym_flags);
1442 trace_print_seq(m, &iter->seq);
bc0c38d1
SR
1443 seq_puts(m, "\n");
1444 }
1445
1446 seq_puts(m, "\n");
1447}
1448
a309720c
SR
1449static void test_cpu_buff_start(struct trace_iterator *iter)
1450{
1451 struct trace_seq *s = &iter->seq;
1452
12ef7d44
SR
1453 if (!(trace_flags & TRACE_ITER_ANNOTATE))
1454 return;
1455
1456 if (!(iter->iter_flags & TRACE_FILE_ANNOTATE))
1457 return;
1458
4462344e 1459 if (cpumask_test_cpu(iter->cpu, iter->started))
a309720c
SR
1460 return;
1461
4462344e 1462 cpumask_set_cpu(iter->cpu, iter->started);
a309720c
SR
1463 trace_seq_printf(s, "##### CPU %u buffer started ####\n", iter->cpu);
1464}
1465
c4a8e8be 1466static enum print_line_t print_lat_fmt(struct trace_iterator *iter)
bc0c38d1 1467{
214023c3 1468 struct trace_seq *s = &iter->seq;
bc0c38d1 1469 unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
f633cef0 1470 struct trace_event *event;
bc0c38d1 1471 struct trace_entry *entry = iter->ent;
dd0e545f 1472
a309720c
SR
1473 test_cpu_buff_start(iter);
1474
c4a8e8be
FW
1475 event = ftrace_find_event(entry->type);
1476
1477 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
d9793bd8
ACM
1478 if (!trace_print_lat_context(iter))
1479 goto partial;
bc0c38d1 1480 }
777e208d 1481
268ccda0 1482 if (event)
d9793bd8
ACM
1483 return event->latency_trace(iter, sym_flags);
1484
1485 if (!trace_seq_printf(s, "Unknown type %d\n", entry->type))
1486 goto partial;
02b67518 1487
2c4f035f 1488 return TRACE_TYPE_HANDLED;
d9793bd8
ACM
1489partial:
1490 return TRACE_TYPE_PARTIAL_LINE;
bc0c38d1
SR
1491}
1492
2c4f035f 1493static enum print_line_t print_trace_fmt(struct trace_iterator *iter)
bc0c38d1 1494{
214023c3 1495 struct trace_seq *s = &iter->seq;
bc0c38d1 1496 unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
4e3c3333 1497 struct trace_entry *entry;
f633cef0 1498 struct trace_event *event;
bc0c38d1 1499
4e3c3333 1500 entry = iter->ent;
dd0e545f 1501
a309720c
SR
1502 test_cpu_buff_start(iter);
1503
c4a8e8be 1504 event = ftrace_find_event(entry->type);
bc0c38d1 1505
c4a8e8be 1506 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
d9793bd8
ACM
1507 if (!trace_print_context(iter))
1508 goto partial;
c4a8e8be 1509 }
bc0c38d1 1510
268ccda0 1511 if (event)
d9793bd8
ACM
1512 return event->trace(iter, sym_flags);
1513
1514 if (!trace_seq_printf(s, "Unknown type %d\n", entry->type))
1515 goto partial;
02b67518 1516
2c4f035f 1517 return TRACE_TYPE_HANDLED;
d9793bd8
ACM
1518partial:
1519 return TRACE_TYPE_PARTIAL_LINE;
bc0c38d1
SR
1520}
1521
2c4f035f 1522static enum print_line_t print_raw_fmt(struct trace_iterator *iter)
f9896bf3
IM
1523{
1524 struct trace_seq *s = &iter->seq;
1525 struct trace_entry *entry;
f633cef0 1526 struct trace_event *event;
f9896bf3
IM
1527
1528 entry = iter->ent;
dd0e545f 1529
c4a8e8be 1530 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
d9793bd8
ACM
1531 if (!trace_seq_printf(s, "%d %d %llu ",
1532 entry->pid, iter->cpu, iter->ts))
1533 goto partial;
c4a8e8be 1534 }
f9896bf3 1535
f633cef0 1536 event = ftrace_find_event(entry->type);
268ccda0 1537 if (event)
d9793bd8
ACM
1538 return event->raw(iter, 0);
1539
1540 if (!trace_seq_printf(s, "%d ?\n", entry->type))
1541 goto partial;
777e208d 1542
2c4f035f 1543 return TRACE_TYPE_HANDLED;
d9793bd8
ACM
1544partial:
1545 return TRACE_TYPE_PARTIAL_LINE;
f9896bf3
IM
1546}
1547
2c4f035f 1548static enum print_line_t print_hex_fmt(struct trace_iterator *iter)
5e3ca0ec
IM
1549{
1550 struct trace_seq *s = &iter->seq;
1551 unsigned char newline = '\n';
1552 struct trace_entry *entry;
f633cef0 1553 struct trace_event *event;
5e3ca0ec
IM
1554
1555 entry = iter->ent;
dd0e545f 1556
c4a8e8be
FW
1557 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
1558 SEQ_PUT_HEX_FIELD_RET(s, entry->pid);
1559 SEQ_PUT_HEX_FIELD_RET(s, iter->cpu);
1560 SEQ_PUT_HEX_FIELD_RET(s, iter->ts);
1561 }
5e3ca0ec 1562
f633cef0 1563 event = ftrace_find_event(entry->type);
268ccda0 1564 if (event) {
ae7462b4 1565 enum print_line_t ret = event->hex(iter, 0);
d9793bd8
ACM
1566 if (ret != TRACE_TYPE_HANDLED)
1567 return ret;
1568 }
7104f300 1569
5e3ca0ec
IM
1570 SEQ_PUT_FIELD_RET(s, newline);
1571
2c4f035f 1572 return TRACE_TYPE_HANDLED;
5e3ca0ec
IM
1573}
1574
66896a85
FW
1575static enum print_line_t print_printk_msg_only(struct trace_iterator *iter)
1576{
1577 struct trace_seq *s = &iter->seq;
1578 struct trace_entry *entry = iter->ent;
1579 struct print_entry *field;
1580 int ret;
1581
1582 trace_assign_type(field, entry);
1583
c4a8e8be 1584 ret = trace_seq_printf(s, "%s", field->buf);
66896a85
FW
1585 if (!ret)
1586 return TRACE_TYPE_PARTIAL_LINE;
1587
66896a85
FW
1588 return TRACE_TYPE_HANDLED;
1589}
1590
2c4f035f 1591static enum print_line_t print_bin_fmt(struct trace_iterator *iter)
cb0f12aa
IM
1592{
1593 struct trace_seq *s = &iter->seq;
1594 struct trace_entry *entry;
f633cef0 1595 struct trace_event *event;
cb0f12aa
IM
1596
1597 entry = iter->ent;
dd0e545f 1598
c4a8e8be
FW
1599 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
1600 SEQ_PUT_FIELD_RET(s, entry->pid);
1830b52d 1601 SEQ_PUT_FIELD_RET(s, iter->cpu);
c4a8e8be
FW
1602 SEQ_PUT_FIELD_RET(s, iter->ts);
1603 }
cb0f12aa 1604
f633cef0 1605 event = ftrace_find_event(entry->type);
268ccda0 1606 return event ? event->binary(iter, 0) : TRACE_TYPE_HANDLED;
cb0f12aa
IM
1607}
1608
bc0c38d1
SR
1609static int trace_empty(struct trace_iterator *iter)
1610{
bc0c38d1
SR
1611 int cpu;
1612
ab46428c 1613 for_each_tracing_cpu(cpu) {
d769041f
SR
1614 if (iter->buffer_iter[cpu]) {
1615 if (!ring_buffer_iter_empty(iter->buffer_iter[cpu]))
1616 return 0;
1617 } else {
1618 if (!ring_buffer_empty_cpu(iter->tr->buffer, cpu))
1619 return 0;
1620 }
bc0c38d1 1621 }
d769041f 1622
797d3712 1623 return 1;
bc0c38d1
SR
1624}
1625
2c4f035f 1626static enum print_line_t print_trace_line(struct trace_iterator *iter)
f9896bf3 1627{
2c4f035f
FW
1628 enum print_line_t ret;
1629
1630 if (iter->trace && iter->trace->print_line) {
1631 ret = iter->trace->print_line(iter);
1632 if (ret != TRACE_TYPE_UNHANDLED)
1633 return ret;
1634 }
72829bc3 1635
66896a85
FW
1636 if (iter->ent->type == TRACE_PRINT &&
1637 trace_flags & TRACE_ITER_PRINTK &&
1638 trace_flags & TRACE_ITER_PRINTK_MSGONLY)
1639 return print_printk_msg_only(iter);
1640
cb0f12aa
IM
1641 if (trace_flags & TRACE_ITER_BIN)
1642 return print_bin_fmt(iter);
1643
5e3ca0ec
IM
1644 if (trace_flags & TRACE_ITER_HEX)
1645 return print_hex_fmt(iter);
1646
f9896bf3
IM
1647 if (trace_flags & TRACE_ITER_RAW)
1648 return print_raw_fmt(iter);
1649
1650 if (iter->iter_flags & TRACE_FILE_LAT_FMT)
c4a8e8be 1651 return print_lat_fmt(iter);
f9896bf3
IM
1652
1653 return print_trace_fmt(iter);
1654}
1655
bc0c38d1
SR
1656static int s_show(struct seq_file *m, void *v)
1657{
1658 struct trace_iterator *iter = v;
1659
1660 if (iter->ent == NULL) {
1661 if (iter->tr) {
1662 seq_printf(m, "# tracer: %s\n", iter->trace->name);
1663 seq_puts(m, "#\n");
1664 }
8bba1bf5
MM
1665 if (iter->trace && iter->trace->print_header)
1666 iter->trace->print_header(m);
1667 else if (iter->iter_flags & TRACE_FILE_LAT_FMT) {
bc0c38d1
SR
1668 /* print nothing if the buffers are empty */
1669 if (trace_empty(iter))
1670 return 0;
1671 print_trace_header(m, iter);
1672 if (!(trace_flags & TRACE_ITER_VERBOSE))
1673 print_lat_help_header(m);
1674 } else {
1675 if (!(trace_flags & TRACE_ITER_VERBOSE))
1676 print_func_help_header(m);
1677 }
1678 } else {
f9896bf3 1679 print_trace_line(iter);
214023c3 1680 trace_print_seq(m, &iter->seq);
bc0c38d1
SR
1681 }
1682
1683 return 0;
1684}
1685
1686static struct seq_operations tracer_seq_ops = {
4bf39a94
IM
1687 .start = s_start,
1688 .next = s_next,
1689 .stop = s_stop,
1690 .show = s_show,
bc0c38d1
SR
1691};
1692
e309b41d 1693static struct trace_iterator *
85a2f9b4 1694__tracing_open(struct inode *inode, struct file *file)
bc0c38d1 1695{
b04cc6b1 1696 long cpu_file = (long) inode->i_private;
85a2f9b4 1697 void *fail_ret = ERR_PTR(-ENOMEM);
bc0c38d1 1698 struct trace_iterator *iter;
3928a8a2 1699 struct seq_file *m;
85a2f9b4 1700 int cpu, ret;
bc0c38d1 1701
85a2f9b4
SR
1702 if (tracing_disabled)
1703 return ERR_PTR(-ENODEV);
60a11774 1704
bc0c38d1 1705 iter = kzalloc(sizeof(*iter), GFP_KERNEL);
85a2f9b4
SR
1706 if (!iter)
1707 return ERR_PTR(-ENOMEM);
bc0c38d1 1708
d7350c3f
FW
1709 /*
1710 * We make a copy of the current tracer to avoid concurrent
1711 * changes on it while we are reading.
1712 */
bc0c38d1 1713 mutex_lock(&trace_types_lock);
d7350c3f 1714 iter->trace = kzalloc(sizeof(*iter->trace), GFP_KERNEL);
85a2f9b4 1715 if (!iter->trace)
d7350c3f 1716 goto fail;
85a2f9b4 1717
d7350c3f
FW
1718 if (current_trace)
1719 *iter->trace = *current_trace;
1720
bc0c38d1
SR
1721 if (current_trace && current_trace->print_max)
1722 iter->tr = &max_tr;
1723 else
b04cc6b1 1724 iter->tr = &global_trace;
bc0c38d1 1725 iter->pos = -1;
d7350c3f 1726 mutex_init(&iter->mutex);
b04cc6b1 1727 iter->cpu_file = cpu_file;
bc0c38d1 1728
8bba1bf5
MM
1729 /* Notify the tracer early; before we stop tracing. */
1730 if (iter->trace && iter->trace->open)
a93751ca 1731 iter->trace->open(iter);
8bba1bf5 1732
12ef7d44
SR
1733 /* Annotate start of buffers if we had overruns */
1734 if (ring_buffer_overruns(iter->tr->buffer))
1735 iter->iter_flags |= TRACE_FILE_ANNOTATE;
1736
b04cc6b1
FW
1737 if (iter->cpu_file == TRACE_PIPE_ALL_CPU) {
1738 for_each_tracing_cpu(cpu) {
12ef7d44 1739
b04cc6b1
FW
1740 iter->buffer_iter[cpu] =
1741 ring_buffer_read_start(iter->tr->buffer, cpu);
d769041f 1742
b04cc6b1
FW
1743 if (!iter->buffer_iter[cpu])
1744 goto fail_buffer;
1745 }
1746 } else {
1747 cpu = iter->cpu_file;
3928a8a2 1748 iter->buffer_iter[cpu] =
b04cc6b1 1749 ring_buffer_read_start(iter->tr->buffer, cpu);
d769041f 1750
3928a8a2 1751 if (!iter->buffer_iter[cpu])
b04cc6b1 1752 goto fail;
3928a8a2
SR
1753 }
1754
bc0c38d1 1755 /* TODO stop tracer */
85a2f9b4
SR
1756 ret = seq_open(file, &tracer_seq_ops);
1757 if (ret < 0) {
1758 fail_ret = ERR_PTR(ret);
3928a8a2 1759 goto fail_buffer;
85a2f9b4 1760 }
bc0c38d1 1761
3928a8a2
SR
1762 m = file->private_data;
1763 m->private = iter;
bc0c38d1 1764
3928a8a2 1765 /* stop the trace while dumping */
9036990d 1766 tracing_stop();
3928a8a2 1767
bc0c38d1
SR
1768 mutex_unlock(&trace_types_lock);
1769
bc0c38d1 1770 return iter;
3928a8a2
SR
1771
1772 fail_buffer:
1773 for_each_tracing_cpu(cpu) {
1774 if (iter->buffer_iter[cpu])
1775 ring_buffer_read_finish(iter->buffer_iter[cpu]);
1776 }
d7350c3f 1777 fail:
3928a8a2 1778 mutex_unlock(&trace_types_lock);
d7350c3f 1779 kfree(iter->trace);
0bb943c7 1780 kfree(iter);
3928a8a2 1781
85a2f9b4 1782 return fail_ret;
bc0c38d1
SR
1783}
1784
1785int tracing_open_generic(struct inode *inode, struct file *filp)
1786{
60a11774
SR
1787 if (tracing_disabled)
1788 return -ENODEV;
1789
bc0c38d1
SR
1790 filp->private_data = inode->i_private;
1791 return 0;
1792}
1793
4fd27358 1794static int tracing_release(struct inode *inode, struct file *file)
bc0c38d1
SR
1795{
1796 struct seq_file *m = (struct seq_file *)file->private_data;
1797 struct trace_iterator *iter = m->private;
3928a8a2 1798 int cpu;
bc0c38d1
SR
1799
1800 mutex_lock(&trace_types_lock);
3928a8a2
SR
1801 for_each_tracing_cpu(cpu) {
1802 if (iter->buffer_iter[cpu])
1803 ring_buffer_read_finish(iter->buffer_iter[cpu]);
1804 }
1805
bc0c38d1
SR
1806 if (iter->trace && iter->trace->close)
1807 iter->trace->close(iter);
1808
1809 /* reenable tracing if it was previously enabled */
9036990d 1810 tracing_start();
bc0c38d1
SR
1811 mutex_unlock(&trace_types_lock);
1812
1813 seq_release(inode, file);
d7350c3f
FW
1814 mutex_destroy(&iter->mutex);
1815 kfree(iter->trace);
bc0c38d1
SR
1816 kfree(iter);
1817 return 0;
1818}
1819
1820static int tracing_open(struct inode *inode, struct file *file)
1821{
85a2f9b4
SR
1822 struct trace_iterator *iter;
1823 int ret = 0;
bc0c38d1 1824
85a2f9b4
SR
1825 iter = __tracing_open(inode, file);
1826 if (IS_ERR(iter))
1827 ret = PTR_ERR(iter);
bc0c38d1
SR
1828
1829 return ret;
1830}
1831
1832static int tracing_lt_open(struct inode *inode, struct file *file)
1833{
1834 struct trace_iterator *iter;
85a2f9b4 1835 int ret = 0;
bc0c38d1 1836
85a2f9b4 1837 iter = __tracing_open(inode, file);
bc0c38d1 1838
85a2f9b4
SR
1839 if (IS_ERR(iter))
1840 ret = PTR_ERR(iter);
1841 else
bc0c38d1
SR
1842 iter->iter_flags |= TRACE_FILE_LAT_FMT;
1843
1844 return ret;
1845}
1846
1847
e309b41d 1848static void *
bc0c38d1
SR
1849t_next(struct seq_file *m, void *v, loff_t *pos)
1850{
1851 struct tracer *t = m->private;
1852
1853 (*pos)++;
1854
1855 if (t)
1856 t = t->next;
1857
1858 m->private = t;
1859
1860 return t;
1861}
1862
1863static void *t_start(struct seq_file *m, loff_t *pos)
1864{
1865 struct tracer *t = m->private;
1866 loff_t l = 0;
1867
1868 mutex_lock(&trace_types_lock);
1869 for (; t && l < *pos; t = t_next(m, t, &l))
1870 ;
1871
1872 return t;
1873}
1874
1875static void t_stop(struct seq_file *m, void *p)
1876{
1877 mutex_unlock(&trace_types_lock);
1878}
1879
1880static int t_show(struct seq_file *m, void *v)
1881{
1882 struct tracer *t = v;
1883
1884 if (!t)
1885 return 0;
1886
1887 seq_printf(m, "%s", t->name);
1888 if (t->next)
1889 seq_putc(m, ' ');
1890 else
1891 seq_putc(m, '\n');
1892
1893 return 0;
1894}
1895
1896static struct seq_operations show_traces_seq_ops = {
4bf39a94
IM
1897 .start = t_start,
1898 .next = t_next,
1899 .stop = t_stop,
1900 .show = t_show,
bc0c38d1
SR
1901};
1902
1903static int show_traces_open(struct inode *inode, struct file *file)
1904{
1905 int ret;
1906
60a11774
SR
1907 if (tracing_disabled)
1908 return -ENODEV;
1909
bc0c38d1
SR
1910 ret = seq_open(file, &show_traces_seq_ops);
1911 if (!ret) {
1912 struct seq_file *m = file->private_data;
1913 m->private = trace_types;
1914 }
1915
1916 return ret;
1917}
1918
1919static struct file_operations tracing_fops = {
4bf39a94
IM
1920 .open = tracing_open,
1921 .read = seq_read,
1922 .llseek = seq_lseek,
1923 .release = tracing_release,
bc0c38d1
SR
1924};
1925
1926static struct file_operations tracing_lt_fops = {
4bf39a94
IM
1927 .open = tracing_lt_open,
1928 .read = seq_read,
1929 .llseek = seq_lseek,
1930 .release = tracing_release,
bc0c38d1
SR
1931};
1932
1933static struct file_operations show_traces_fops = {
c7078de1
IM
1934 .open = show_traces_open,
1935 .read = seq_read,
1936 .release = seq_release,
1937};
1938
36dfe925
IM
1939/*
1940 * Only trace on a CPU if the bitmask is set:
1941 */
9e01c1b7 1942static cpumask_var_t tracing_cpumask;
36dfe925
IM
1943
1944/*
1945 * The tracer itself will not take this lock, but still we want
1946 * to provide a consistent cpumask to user-space:
1947 */
1948static DEFINE_MUTEX(tracing_cpumask_update_lock);
1949
1950/*
1951 * Temporary storage for the character representation of the
1952 * CPU bitmask (and one more byte for the newline):
1953 */
1954static char mask_str[NR_CPUS + 1];
1955
c7078de1
IM
1956static ssize_t
1957tracing_cpumask_read(struct file *filp, char __user *ubuf,
1958 size_t count, loff_t *ppos)
1959{
36dfe925 1960 int len;
c7078de1
IM
1961
1962 mutex_lock(&tracing_cpumask_update_lock);
36dfe925 1963
9e01c1b7 1964 len = cpumask_scnprintf(mask_str, count, tracing_cpumask);
36dfe925
IM
1965 if (count - len < 2) {
1966 count = -EINVAL;
1967 goto out_err;
1968 }
1969 len += sprintf(mask_str + len, "\n");
1970 count = simple_read_from_buffer(ubuf, count, ppos, mask_str, NR_CPUS+1);
1971
1972out_err:
c7078de1
IM
1973 mutex_unlock(&tracing_cpumask_update_lock);
1974
1975 return count;
1976}
1977
1978static ssize_t
1979tracing_cpumask_write(struct file *filp, const char __user *ubuf,
1980 size_t count, loff_t *ppos)
1981{
36dfe925 1982 int err, cpu;
9e01c1b7
RR
1983 cpumask_var_t tracing_cpumask_new;
1984
1985 if (!alloc_cpumask_var(&tracing_cpumask_new, GFP_KERNEL))
1986 return -ENOMEM;
c7078de1
IM
1987
1988 mutex_lock(&tracing_cpumask_update_lock);
9e01c1b7 1989 err = cpumask_parse_user(ubuf, count, tracing_cpumask_new);
c7078de1 1990 if (err)
36dfe925
IM
1991 goto err_unlock;
1992
a5e25883 1993 local_irq_disable();
92205c23 1994 __raw_spin_lock(&ftrace_max_lock);
ab46428c 1995 for_each_tracing_cpu(cpu) {
36dfe925
IM
1996 /*
1997 * Increase/decrease the disabled counter if we are
1998 * about to flip a bit in the cpumask:
1999 */
9e01c1b7
RR
2000 if (cpumask_test_cpu(cpu, tracing_cpumask) &&
2001 !cpumask_test_cpu(cpu, tracing_cpumask_new)) {
36dfe925
IM
2002 atomic_inc(&global_trace.data[cpu]->disabled);
2003 }
9e01c1b7
RR
2004 if (!cpumask_test_cpu(cpu, tracing_cpumask) &&
2005 cpumask_test_cpu(cpu, tracing_cpumask_new)) {
36dfe925
IM
2006 atomic_dec(&global_trace.data[cpu]->disabled);
2007 }
2008 }
92205c23 2009 __raw_spin_unlock(&ftrace_max_lock);
a5e25883 2010 local_irq_enable();
36dfe925 2011
9e01c1b7 2012 cpumask_copy(tracing_cpumask, tracing_cpumask_new);
36dfe925
IM
2013
2014 mutex_unlock(&tracing_cpumask_update_lock);
9e01c1b7 2015 free_cpumask_var(tracing_cpumask_new);
c7078de1
IM
2016
2017 return count;
36dfe925
IM
2018
2019err_unlock:
2020 mutex_unlock(&tracing_cpumask_update_lock);
9e01c1b7 2021 free_cpumask_var(tracing_cpumask);
36dfe925
IM
2022
2023 return err;
c7078de1
IM
2024}
2025
2026static struct file_operations tracing_cpumask_fops = {
2027 .open = tracing_open_generic,
2028 .read = tracing_cpumask_read,
2029 .write = tracing_cpumask_write,
bc0c38d1
SR
2030};
2031
2032static ssize_t
ee6bce52 2033tracing_trace_options_read(struct file *filp, char __user *ubuf,
bc0c38d1
SR
2034 size_t cnt, loff_t *ppos)
2035{
d8e83d26
SR
2036 struct tracer_opt *trace_opts;
2037 u32 tracer_flags;
2038 int len = 0;
bc0c38d1
SR
2039 char *buf;
2040 int r = 0;
d8e83d26 2041 int i;
adf9f195 2042
bc0c38d1 2043
c3706f00 2044 /* calculate max size */
bc0c38d1
SR
2045 for (i = 0; trace_options[i]; i++) {
2046 len += strlen(trace_options[i]);
5c6a3ae1 2047 len += 3; /* "no" and newline */
bc0c38d1
SR
2048 }
2049
d8e83d26
SR
2050 mutex_lock(&trace_types_lock);
2051 tracer_flags = current_trace->flags->val;
2052 trace_opts = current_trace->flags->opts;
2053
adf9f195
FW
2054 /*
2055 * Increase the size with names of options specific
2056 * of the current tracer.
2057 */
2058 for (i = 0; trace_opts[i].name; i++) {
2059 len += strlen(trace_opts[i].name);
5c6a3ae1 2060 len += 3; /* "no" and newline */
adf9f195
FW
2061 }
2062
bc0c38d1
SR
2063 /* +2 for \n and \0 */
2064 buf = kmalloc(len + 2, GFP_KERNEL);
d8e83d26
SR
2065 if (!buf) {
2066 mutex_unlock(&trace_types_lock);
bc0c38d1 2067 return -ENOMEM;
d8e83d26 2068 }
bc0c38d1
SR
2069
2070 for (i = 0; trace_options[i]; i++) {
2071 if (trace_flags & (1 << i))
5c6a3ae1 2072 r += sprintf(buf + r, "%s\n", trace_options[i]);
bc0c38d1 2073 else
5c6a3ae1 2074 r += sprintf(buf + r, "no%s\n", trace_options[i]);
bc0c38d1
SR
2075 }
2076
adf9f195
FW
2077 for (i = 0; trace_opts[i].name; i++) {
2078 if (tracer_flags & trace_opts[i].bit)
5c6a3ae1 2079 r += sprintf(buf + r, "%s\n",
adf9f195
FW
2080 trace_opts[i].name);
2081 else
5c6a3ae1 2082 r += sprintf(buf + r, "no%s\n",
adf9f195
FW
2083 trace_opts[i].name);
2084 }
d8e83d26 2085 mutex_unlock(&trace_types_lock);
adf9f195 2086
bc0c38d1
SR
2087 WARN_ON(r >= len + 2);
2088
36dfe925 2089 r = simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
bc0c38d1
SR
2090
2091 kfree(buf);
bc0c38d1
SR
2092 return r;
2093}
2094
adf9f195
FW
2095/* Try to assign a tracer specific option */
2096static int set_tracer_option(struct tracer *trace, char *cmp, int neg)
2097{
2098 struct tracer_flags *trace_flags = trace->flags;
2099 struct tracer_opt *opts = NULL;
2100 int ret = 0, i = 0;
2101 int len;
2102
2103 for (i = 0; trace_flags->opts[i].name; i++) {
2104 opts = &trace_flags->opts[i];
2105 len = strlen(opts->name);
2106
2107 if (strncmp(cmp, opts->name, len) == 0) {
2108 ret = trace->set_flag(trace_flags->val,
2109 opts->bit, !neg);
2110 break;
2111 }
2112 }
2113 /* Not found */
2114 if (!trace_flags->opts[i].name)
2115 return -EINVAL;
2116
2117 /* Refused to handle */
2118 if (ret)
2119 return ret;
2120
2121 if (neg)
2122 trace_flags->val &= ~opts->bit;
2123 else
2124 trace_flags->val |= opts->bit;
2125
2126 return 0;
2127}
2128
bc0c38d1 2129static ssize_t
ee6bce52 2130tracing_trace_options_write(struct file *filp, const char __user *ubuf,
bc0c38d1
SR
2131 size_t cnt, loff_t *ppos)
2132{
2133 char buf[64];
2134 char *cmp = buf;
2135 int neg = 0;
adf9f195 2136 int ret;
bc0c38d1
SR
2137 int i;
2138
cffae437
SR
2139 if (cnt >= sizeof(buf))
2140 return -EINVAL;
bc0c38d1
SR
2141
2142 if (copy_from_user(&buf, ubuf, cnt))
2143 return -EFAULT;
2144
2145 buf[cnt] = 0;
2146
2147 if (strncmp(buf, "no", 2) == 0) {
2148 neg = 1;
2149 cmp += 2;
2150 }
2151
2152 for (i = 0; trace_options[i]; i++) {
2153 int len = strlen(trace_options[i]);
2154
2155 if (strncmp(cmp, trace_options[i], len) == 0) {
2156 if (neg)
2157 trace_flags &= ~(1 << i);
2158 else
2159 trace_flags |= (1 << i);
2160 break;
2161 }
2162 }
adf9f195
FW
2163
2164 /* If no option could be set, test the specific tracer options */
2165 if (!trace_options[i]) {
d8e83d26 2166 mutex_lock(&trace_types_lock);
adf9f195 2167 ret = set_tracer_option(current_trace, cmp, neg);
d8e83d26 2168 mutex_unlock(&trace_types_lock);
adf9f195
FW
2169 if (ret)
2170 return ret;
2171 }
bc0c38d1
SR
2172
2173 filp->f_pos += cnt;
2174
2175 return cnt;
2176}
2177
2178static struct file_operations tracing_iter_fops = {
c7078de1 2179 .open = tracing_open_generic,
ee6bce52
SR
2180 .read = tracing_trace_options_read,
2181 .write = tracing_trace_options_write,
bc0c38d1
SR
2182};
2183
7bd2f24c
IM
2184static const char readme_msg[] =
2185 "tracing mini-HOWTO:\n\n"
2186 "# mkdir /debug\n"
2187 "# mount -t debugfs nodev /debug\n\n"
2188 "# cat /debug/tracing/available_tracers\n"
2189 "wakeup preemptirqsoff preemptoff irqsoff ftrace sched_switch none\n\n"
2190 "# cat /debug/tracing/current_tracer\n"
2191 "none\n"
2192 "# echo sched_switch > /debug/tracing/current_tracer\n"
2193 "# cat /debug/tracing/current_tracer\n"
2194 "sched_switch\n"
ee6bce52 2195 "# cat /debug/tracing/trace_options\n"
7bd2f24c 2196 "noprint-parent nosym-offset nosym-addr noverbose\n"
ee6bce52 2197 "# echo print-parent > /debug/tracing/trace_options\n"
7bd2f24c
IM
2198 "# echo 1 > /debug/tracing/tracing_enabled\n"
2199 "# cat /debug/tracing/trace > /tmp/trace.txt\n"
2200 "echo 0 > /debug/tracing/tracing_enabled\n"
2201;
2202
2203static ssize_t
2204tracing_readme_read(struct file *filp, char __user *ubuf,
2205 size_t cnt, loff_t *ppos)
2206{
2207 return simple_read_from_buffer(ubuf, cnt, ppos,
2208 readme_msg, strlen(readme_msg));
2209}
2210
2211static struct file_operations tracing_readme_fops = {
c7078de1
IM
2212 .open = tracing_open_generic,
2213 .read = tracing_readme_read,
7bd2f24c
IM
2214};
2215
bc0c38d1
SR
2216static ssize_t
2217tracing_ctrl_read(struct file *filp, char __user *ubuf,
2218 size_t cnt, loff_t *ppos)
2219{
bc0c38d1
SR
2220 char buf[64];
2221 int r;
2222
9036990d 2223 r = sprintf(buf, "%u\n", tracer_enabled);
4e3c3333 2224 return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
bc0c38d1
SR
2225}
2226
2227static ssize_t
2228tracing_ctrl_write(struct file *filp, const char __user *ubuf,
2229 size_t cnt, loff_t *ppos)
2230{
2231 struct trace_array *tr = filp->private_data;
bc0c38d1 2232 char buf[64];
5e39841c 2233 unsigned long val;
c6caeeb1 2234 int ret;
bc0c38d1 2235
cffae437
SR
2236 if (cnt >= sizeof(buf))
2237 return -EINVAL;
bc0c38d1
SR
2238
2239 if (copy_from_user(&buf, ubuf, cnt))
2240 return -EFAULT;
2241
2242 buf[cnt] = 0;
2243
c6caeeb1
SR
2244 ret = strict_strtoul(buf, 10, &val);
2245 if (ret < 0)
2246 return ret;
bc0c38d1
SR
2247
2248 val = !!val;
2249
2250 mutex_lock(&trace_types_lock);
9036990d
SR
2251 if (tracer_enabled ^ val) {
2252 if (val) {
bc0c38d1 2253 tracer_enabled = 1;
9036990d
SR
2254 if (current_trace->start)
2255 current_trace->start(tr);
2256 tracing_start();
2257 } else {
bc0c38d1 2258 tracer_enabled = 0;
9036990d
SR
2259 tracing_stop();
2260 if (current_trace->stop)
2261 current_trace->stop(tr);
2262 }
bc0c38d1
SR
2263 }
2264 mutex_unlock(&trace_types_lock);
2265
2266 filp->f_pos += cnt;
2267
2268 return cnt;
2269}
2270
2271static ssize_t
2272tracing_set_trace_read(struct file *filp, char __user *ubuf,
2273 size_t cnt, loff_t *ppos)
2274{
2275 char buf[max_tracer_type_len+2];
2276 int r;
2277
2278 mutex_lock(&trace_types_lock);
2279 if (current_trace)
2280 r = sprintf(buf, "%s\n", current_trace->name);
2281 else
2282 r = sprintf(buf, "\n");
2283 mutex_unlock(&trace_types_lock);
2284
4bf39a94 2285 return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
bc0c38d1
SR
2286}
2287
b6f11df2
ACM
2288int tracer_init(struct tracer *t, struct trace_array *tr)
2289{
2290 tracing_reset_online_cpus(tr);
2291 return t->init(tr);
2292}
2293
577b785f
SR
2294struct trace_option_dentry;
2295
2296static struct trace_option_dentry *
2297create_trace_option_files(struct tracer *tracer);
2298
2299static void
2300destroy_trace_option_files(struct trace_option_dentry *topts);
2301
b2821ae6 2302static int tracing_set_tracer(const char *buf)
bc0c38d1 2303{
577b785f 2304 static struct trace_option_dentry *topts;
bc0c38d1
SR
2305 struct trace_array *tr = &global_trace;
2306 struct tracer *t;
d9e54076 2307 int ret = 0;
bc0c38d1
SR
2308
2309 mutex_lock(&trace_types_lock);
2310 for (t = trace_types; t; t = t->next) {
2311 if (strcmp(t->name, buf) == 0)
2312 break;
2313 }
c2931e05
FW
2314 if (!t) {
2315 ret = -EINVAL;
2316 goto out;
2317 }
2318 if (t == current_trace)
bc0c38d1
SR
2319 goto out;
2320
9f029e83 2321 trace_branch_disable();
bc0c38d1
SR
2322 if (current_trace && current_trace->reset)
2323 current_trace->reset(tr);
2324
577b785f
SR
2325 destroy_trace_option_files(topts);
2326
bc0c38d1 2327 current_trace = t;
577b785f
SR
2328
2329 topts = create_trace_option_files(current_trace);
2330
1c80025a 2331 if (t->init) {
b6f11df2 2332 ret = tracer_init(t, tr);
1c80025a
FW
2333 if (ret)
2334 goto out;
2335 }
bc0c38d1 2336
9f029e83 2337 trace_branch_enable(tr);
bc0c38d1
SR
2338 out:
2339 mutex_unlock(&trace_types_lock);
2340
d9e54076
PZ
2341 return ret;
2342}
2343
2344static ssize_t
2345tracing_set_trace_write(struct file *filp, const char __user *ubuf,
2346 size_t cnt, loff_t *ppos)
2347{
2348 char buf[max_tracer_type_len+1];
2349 int i;
2350 size_t ret;
e6e7a65a
FW
2351 int err;
2352
2353 ret = cnt;
d9e54076
PZ
2354
2355 if (cnt > max_tracer_type_len)
2356 cnt = max_tracer_type_len;
2357
2358 if (copy_from_user(&buf, ubuf, cnt))
2359 return -EFAULT;
2360
2361 buf[cnt] = 0;
2362
2363 /* strip ending whitespace. */
2364 for (i = cnt - 1; i > 0 && isspace(buf[i]); i--)
2365 buf[i] = 0;
2366
e6e7a65a
FW
2367 err = tracing_set_tracer(buf);
2368 if (err)
2369 return err;
d9e54076 2370
e6e7a65a 2371 filp->f_pos += ret;
bc0c38d1 2372
c2931e05 2373 return ret;
bc0c38d1
SR
2374}
2375
2376static ssize_t
2377tracing_max_lat_read(struct file *filp, char __user *ubuf,
2378 size_t cnt, loff_t *ppos)
2379{
2380 unsigned long *ptr = filp->private_data;
2381 char buf[64];
2382 int r;
2383
cffae437 2384 r = snprintf(buf, sizeof(buf), "%ld\n",
bc0c38d1 2385 *ptr == (unsigned long)-1 ? -1 : nsecs_to_usecs(*ptr));
cffae437
SR
2386 if (r > sizeof(buf))
2387 r = sizeof(buf);
4bf39a94 2388 return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
bc0c38d1
SR
2389}
2390
2391static ssize_t
2392tracing_max_lat_write(struct file *filp, const char __user *ubuf,
2393 size_t cnt, loff_t *ppos)
2394{
5e39841c 2395 unsigned long *ptr = filp->private_data;
bc0c38d1 2396 char buf[64];
5e39841c 2397 unsigned long val;
c6caeeb1 2398 int ret;
bc0c38d1 2399
cffae437
SR
2400 if (cnt >= sizeof(buf))
2401 return -EINVAL;
bc0c38d1
SR
2402
2403 if (copy_from_user(&buf, ubuf, cnt))
2404 return -EFAULT;
2405
2406 buf[cnt] = 0;
2407
c6caeeb1
SR
2408 ret = strict_strtoul(buf, 10, &val);
2409 if (ret < 0)
2410 return ret;
bc0c38d1
SR
2411
2412 *ptr = val * 1000;
2413
2414 return cnt;
2415}
2416
b3806b43
SR
2417static int tracing_open_pipe(struct inode *inode, struct file *filp)
2418{
b04cc6b1 2419 long cpu_file = (long) inode->i_private;
b3806b43 2420 struct trace_iterator *iter;
b04cc6b1 2421 int ret = 0;
b3806b43
SR
2422
2423 if (tracing_disabled)
2424 return -ENODEV;
2425
b04cc6b1
FW
2426 mutex_lock(&trace_types_lock);
2427
2428 /* We only allow one reader per cpu */
2429 if (cpu_file == TRACE_PIPE_ALL_CPU) {
2430 if (!cpumask_empty(tracing_reader_cpumask)) {
2431 ret = -EBUSY;
2432 goto out;
2433 }
2434 cpumask_setall(tracing_reader_cpumask);
2435 } else {
2436 if (!cpumask_test_cpu(cpu_file, tracing_reader_cpumask))
2437 cpumask_set_cpu(cpu_file, tracing_reader_cpumask);
2438 else {
2439 ret = -EBUSY;
2440 goto out;
2441 }
b3806b43
SR
2442 }
2443
2444 /* create a buffer to store the information to pass to userspace */
2445 iter = kzalloc(sizeof(*iter), GFP_KERNEL);
b04cc6b1
FW
2446 if (!iter) {
2447 ret = -ENOMEM;
2448 goto out;
2449 }
b3806b43 2450
d7350c3f
FW
2451 /*
2452 * We make a copy of the current tracer to avoid concurrent
2453 * changes on it while we are reading.
2454 */
2455 iter->trace = kmalloc(sizeof(*iter->trace), GFP_KERNEL);
2456 if (!iter->trace) {
2457 ret = -ENOMEM;
2458 goto fail;
2459 }
2460 if (current_trace)
2461 *iter->trace = *current_trace;
2462
4462344e 2463 if (!alloc_cpumask_var(&iter->started, GFP_KERNEL)) {
b04cc6b1 2464 ret = -ENOMEM;
d7350c3f 2465 goto fail;
4462344e
RR
2466 }
2467
a309720c 2468 /* trace pipe does not show start of buffer */
4462344e 2469 cpumask_setall(iter->started);
a309720c 2470
b04cc6b1 2471 iter->cpu_file = cpu_file;
b3806b43 2472 iter->tr = &global_trace;
d7350c3f 2473 mutex_init(&iter->mutex);
b3806b43
SR
2474 filp->private_data = iter;
2475
107bad8b
SR
2476 if (iter->trace->pipe_open)
2477 iter->trace->pipe_open(iter);
107bad8b 2478
b04cc6b1
FW
2479out:
2480 mutex_unlock(&trace_types_lock);
2481 return ret;
d7350c3f
FW
2482
2483fail:
2484 kfree(iter->trace);
2485 kfree(iter);
2486 mutex_unlock(&trace_types_lock);
2487 return ret;
b3806b43
SR
2488}
2489
2490static int tracing_release_pipe(struct inode *inode, struct file *file)
2491{
2492 struct trace_iterator *iter = file->private_data;
2493
b04cc6b1
FW
2494 mutex_lock(&trace_types_lock);
2495
2496 if (iter->cpu_file == TRACE_PIPE_ALL_CPU)
2497 cpumask_clear(tracing_reader_cpumask);
2498 else
2499 cpumask_clear_cpu(iter->cpu_file, tracing_reader_cpumask);
2500
2501 mutex_unlock(&trace_types_lock);
2502
4462344e 2503 free_cpumask_var(iter->started);
d7350c3f
FW
2504 mutex_destroy(&iter->mutex);
2505 kfree(iter->trace);
b3806b43 2506 kfree(iter);
b3806b43
SR
2507
2508 return 0;
2509}
2510
2a2cc8f7
SSP
2511static unsigned int
2512tracing_poll_pipe(struct file *filp, poll_table *poll_table)
2513{
2514 struct trace_iterator *iter = filp->private_data;
2515
2516 if (trace_flags & TRACE_ITER_BLOCK) {
2517 /*
2518 * Always select as readable when in blocking mode
2519 */
2520 return POLLIN | POLLRDNORM;
afc2abc0 2521 } else {
2a2cc8f7
SSP
2522 if (!trace_empty(iter))
2523 return POLLIN | POLLRDNORM;
2524 poll_wait(filp, &trace_wait, poll_table);
2525 if (!trace_empty(iter))
2526 return POLLIN | POLLRDNORM;
2527
2528 return 0;
2529 }
2530}
2531
6eaaa5d5
FW
2532
2533void default_wait_pipe(struct trace_iterator *iter)
2534{
2535 DEFINE_WAIT(wait);
2536
2537 prepare_to_wait(&trace_wait, &wait, TASK_INTERRUPTIBLE);
2538
2539 if (trace_empty(iter))
2540 schedule();
2541
2542 finish_wait(&trace_wait, &wait);
2543}
2544
2545/*
2546 * This is a make-shift waitqueue.
2547 * A tracer might use this callback on some rare cases:
2548 *
2549 * 1) the current tracer might hold the runqueue lock when it wakes up
2550 * a reader, hence a deadlock (sched, function, and function graph tracers)
2551 * 2) the function tracers, trace all functions, we don't want
2552 * the overhead of calling wake_up and friends
2553 * (and tracing them too)
2554 *
2555 * Anyway, this is really very primitive wakeup.
2556 */
2557void poll_wait_pipe(struct trace_iterator *iter)
2558{
2559 set_current_state(TASK_INTERRUPTIBLE);
2560 /* sleep for 100 msecs, and try again. */
2561 schedule_timeout(HZ / 10);
2562}
2563
ff98781b
EGM
2564/* Must be called with trace_types_lock mutex held. */
2565static int tracing_wait_pipe(struct file *filp)
b3806b43
SR
2566{
2567 struct trace_iterator *iter = filp->private_data;
b3806b43 2568
b3806b43 2569 while (trace_empty(iter)) {
2dc8f095 2570
107bad8b 2571 if ((filp->f_flags & O_NONBLOCK)) {
ff98781b 2572 return -EAGAIN;
107bad8b 2573 }
2dc8f095 2574
d7350c3f 2575 mutex_unlock(&iter->mutex);
107bad8b 2576
6eaaa5d5 2577 iter->trace->wait_pipe(iter);
b3806b43 2578
d7350c3f 2579 mutex_lock(&iter->mutex);
107bad8b 2580
6eaaa5d5 2581 if (signal_pending(current))
ff98781b 2582 return -EINTR;
b3806b43
SR
2583
2584 /*
2585 * We block until we read something and tracing is disabled.
2586 * We still block if tracing is disabled, but we have never
2587 * read anything. This allows a user to cat this file, and
2588 * then enable tracing. But after we have read something,
2589 * we give an EOF when tracing is again disabled.
2590 *
2591 * iter->pos will be 0 if we haven't read anything.
2592 */
2593 if (!tracer_enabled && iter->pos)
2594 break;
b3806b43
SR
2595 }
2596
ff98781b
EGM
2597 return 1;
2598}
2599
2600/*
2601 * Consumer reader.
2602 */
2603static ssize_t
2604tracing_read_pipe(struct file *filp, char __user *ubuf,
2605 size_t cnt, loff_t *ppos)
2606{
2607 struct trace_iterator *iter = filp->private_data;
d7350c3f 2608 static struct tracer *old_tracer;
ff98781b
EGM
2609 ssize_t sret;
2610
2611 /* return any leftover data */
2612 sret = trace_seq_to_user(&iter->seq, ubuf, cnt);
2613 if (sret != -EBUSY)
2614 return sret;
2615
f9520750 2616 trace_seq_init(&iter->seq);
ff98781b 2617
d7350c3f 2618 /* copy the tracer to avoid using a global lock all around */
ff98781b 2619 mutex_lock(&trace_types_lock);
d7350c3f
FW
2620 if (unlikely(old_tracer != current_trace && current_trace)) {
2621 old_tracer = current_trace;
2622 *iter->trace = *current_trace;
2623 }
2624 mutex_unlock(&trace_types_lock);
2625
2626 /*
2627 * Avoid more than one consumer on a single file descriptor
2628 * This is just a matter of traces coherency, the ring buffer itself
2629 * is protected.
2630 */
2631 mutex_lock(&iter->mutex);
ff98781b
EGM
2632 if (iter->trace->read) {
2633 sret = iter->trace->read(iter, filp, ubuf, cnt, ppos);
2634 if (sret)
2635 goto out;
2636 }
2637
2638waitagain:
2639 sret = tracing_wait_pipe(filp);
2640 if (sret <= 0)
2641 goto out;
2642
b3806b43 2643 /* stop when tracing is finished */
ff98781b
EGM
2644 if (trace_empty(iter)) {
2645 sret = 0;
107bad8b 2646 goto out;
ff98781b 2647 }
b3806b43
SR
2648
2649 if (cnt >= PAGE_SIZE)
2650 cnt = PAGE_SIZE - 1;
2651
53d0aa77 2652 /* reset all but tr, trace, and overruns */
53d0aa77
SR
2653 memset(&iter->seq, 0,
2654 sizeof(struct trace_iterator) -
2655 offsetof(struct trace_iterator, seq));
4823ed7e 2656 iter->pos = -1;
b3806b43 2657
088b1e42 2658 while (find_next_entry_inc(iter) != NULL) {
2c4f035f 2659 enum print_line_t ret;
088b1e42
SR
2660 int len = iter->seq.len;
2661
f9896bf3 2662 ret = print_trace_line(iter);
2c4f035f 2663 if (ret == TRACE_TYPE_PARTIAL_LINE) {
088b1e42
SR
2664 /* don't print partial lines */
2665 iter->seq.len = len;
b3806b43 2666 break;
088b1e42 2667 }
b91facc3
FW
2668 if (ret != TRACE_TYPE_NO_CONSUME)
2669 trace_consume(iter);
b3806b43
SR
2670
2671 if (iter->seq.len >= cnt)
2672 break;
b3806b43
SR
2673 }
2674
b3806b43 2675 /* Now copy what we have to the user */
6c6c2796
PP
2676 sret = trace_seq_to_user(&iter->seq, ubuf, cnt);
2677 if (iter->seq.readpos >= iter->seq.len)
f9520750 2678 trace_seq_init(&iter->seq);
9ff4b974
PP
2679
2680 /*
2681 * If there was nothing to send to user, inspite of consuming trace
2682 * entries, go back to wait for more entries.
2683 */
6c6c2796 2684 if (sret == -EBUSY)
9ff4b974 2685 goto waitagain;
b3806b43 2686
107bad8b 2687out:
d7350c3f 2688 mutex_unlock(&iter->mutex);
107bad8b 2689
6c6c2796 2690 return sret;
b3806b43
SR
2691}
2692
3c56819b
EGM
2693static void tracing_pipe_buf_release(struct pipe_inode_info *pipe,
2694 struct pipe_buffer *buf)
2695{
2696 __free_page(buf->page);
2697}
2698
2699static void tracing_spd_release_pipe(struct splice_pipe_desc *spd,
2700 unsigned int idx)
2701{
2702 __free_page(spd->pages[idx]);
2703}
2704
2705static struct pipe_buf_operations tracing_pipe_buf_ops = {
34cd4998
SR
2706 .can_merge = 0,
2707 .map = generic_pipe_buf_map,
2708 .unmap = generic_pipe_buf_unmap,
2709 .confirm = generic_pipe_buf_confirm,
2710 .release = tracing_pipe_buf_release,
2711 .steal = generic_pipe_buf_steal,
2712 .get = generic_pipe_buf_get,
3c56819b
EGM
2713};
2714
34cd4998 2715static size_t
fa7c7f6e 2716tracing_fill_pipe_page(size_t rem, struct trace_iterator *iter)
34cd4998
SR
2717{
2718 size_t count;
2719 int ret;
2720
2721 /* Seq buffer is page-sized, exactly what we need. */
2722 for (;;) {
2723 count = iter->seq.len;
2724 ret = print_trace_line(iter);
2725 count = iter->seq.len - count;
2726 if (rem < count) {
2727 rem = 0;
2728 iter->seq.len -= count;
2729 break;
2730 }
2731 if (ret == TRACE_TYPE_PARTIAL_LINE) {
2732 iter->seq.len -= count;
2733 break;
2734 }
2735
2736 trace_consume(iter);
2737 rem -= count;
2738 if (!find_next_entry_inc(iter)) {
2739 rem = 0;
2740 iter->ent = NULL;
2741 break;
2742 }
2743 }
2744
2745 return rem;
2746}
2747
3c56819b
EGM
2748static ssize_t tracing_splice_read_pipe(struct file *filp,
2749 loff_t *ppos,
2750 struct pipe_inode_info *pipe,
2751 size_t len,
2752 unsigned int flags)
2753{
2754 struct page *pages[PIPE_BUFFERS];
2755 struct partial_page partial[PIPE_BUFFERS];
2756 struct trace_iterator *iter = filp->private_data;
2757 struct splice_pipe_desc spd = {
34cd4998
SR
2758 .pages = pages,
2759 .partial = partial,
2760 .nr_pages = 0, /* This gets updated below. */
2761 .flags = flags,
2762 .ops = &tracing_pipe_buf_ops,
2763 .spd_release = tracing_spd_release_pipe,
3c56819b 2764 };
d7350c3f 2765 static struct tracer *old_tracer;
3c56819b 2766 ssize_t ret;
34cd4998 2767 size_t rem;
3c56819b
EGM
2768 unsigned int i;
2769
d7350c3f 2770 /* copy the tracer to avoid using a global lock all around */
3c56819b 2771 mutex_lock(&trace_types_lock);
d7350c3f
FW
2772 if (unlikely(old_tracer != current_trace && current_trace)) {
2773 old_tracer = current_trace;
2774 *iter->trace = *current_trace;
2775 }
2776 mutex_unlock(&trace_types_lock);
2777
2778 mutex_lock(&iter->mutex);
3c56819b
EGM
2779
2780 if (iter->trace->splice_read) {
2781 ret = iter->trace->splice_read(iter, filp,
2782 ppos, pipe, len, flags);
2783 if (ret)
34cd4998 2784 goto out_err;
3c56819b
EGM
2785 }
2786
2787 ret = tracing_wait_pipe(filp);
2788 if (ret <= 0)
34cd4998 2789 goto out_err;
3c56819b
EGM
2790
2791 if (!iter->ent && !find_next_entry_inc(iter)) {
2792 ret = -EFAULT;
34cd4998 2793 goto out_err;
3c56819b
EGM
2794 }
2795
2796 /* Fill as many pages as possible. */
2797 for (i = 0, rem = len; i < PIPE_BUFFERS && rem; i++) {
2798 pages[i] = alloc_page(GFP_KERNEL);
34cd4998
SR
2799 if (!pages[i])
2800 break;
3c56819b 2801
fa7c7f6e 2802 rem = tracing_fill_pipe_page(rem, iter);
3c56819b
EGM
2803
2804 /* Copy the data into the page, so we can start over. */
2805 ret = trace_seq_to_buffer(&iter->seq,
2806 page_address(pages[i]),
2807 iter->seq.len);
2808 if (ret < 0) {
2809 __free_page(pages[i]);
2810 break;
2811 }
2812 partial[i].offset = 0;
2813 partial[i].len = iter->seq.len;
2814
f9520750 2815 trace_seq_init(&iter->seq);
3c56819b
EGM
2816 }
2817
d7350c3f 2818 mutex_unlock(&iter->mutex);
3c56819b
EGM
2819
2820 spd.nr_pages = i;
2821
2822 return splice_to_pipe(pipe, &spd);
2823
34cd4998 2824out_err:
d7350c3f 2825 mutex_unlock(&iter->mutex);
3c56819b
EGM
2826
2827 return ret;
2828}
2829
a98a3c3f
SR
2830static ssize_t
2831tracing_entries_read(struct file *filp, char __user *ubuf,
2832 size_t cnt, loff_t *ppos)
2833{
2834 struct trace_array *tr = filp->private_data;
2835 char buf[64];
2836 int r;
2837
1696b2b0 2838 r = sprintf(buf, "%lu\n", tr->entries >> 10);
a98a3c3f
SR
2839 return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
2840}
2841
2842static ssize_t
2843tracing_entries_write(struct file *filp, const char __user *ubuf,
2844 size_t cnt, loff_t *ppos)
2845{
2846 unsigned long val;
2847 char buf[64];
bf5e6519 2848 int ret, cpu;
a98a3c3f 2849
cffae437
SR
2850 if (cnt >= sizeof(buf))
2851 return -EINVAL;
a98a3c3f
SR
2852
2853 if (copy_from_user(&buf, ubuf, cnt))
2854 return -EFAULT;
2855
2856 buf[cnt] = 0;
2857
c6caeeb1
SR
2858 ret = strict_strtoul(buf, 10, &val);
2859 if (ret < 0)
2860 return ret;
a98a3c3f
SR
2861
2862 /* must have at least 1 entry */
2863 if (!val)
2864 return -EINVAL;
2865
2866 mutex_lock(&trace_types_lock);
2867
c76f0694 2868 tracing_stop();
a98a3c3f 2869
bf5e6519
SR
2870 /* disable all cpu buffers */
2871 for_each_tracing_cpu(cpu) {
2872 if (global_trace.data[cpu])
2873 atomic_inc(&global_trace.data[cpu]->disabled);
2874 if (max_tr.data[cpu])
2875 atomic_inc(&max_tr.data[cpu]->disabled);
2876 }
2877
1696b2b0
SR
2878 /* value is in KB */
2879 val <<= 10;
2880
3928a8a2
SR
2881 if (val != global_trace.entries) {
2882 ret = ring_buffer_resize(global_trace.buffer, val);
2883 if (ret < 0) {
2884 cnt = ret;
3eefae99
SR
2885 goto out;
2886 }
2887
3928a8a2
SR
2888 ret = ring_buffer_resize(max_tr.buffer, val);
2889 if (ret < 0) {
2890 int r;
2891 cnt = ret;
2892 r = ring_buffer_resize(global_trace.buffer,
2893 global_trace.entries);
2894 if (r < 0) {
2895 /* AARGH! We are left with different
2896 * size max buffer!!!! */
2897 WARN_ON(1);
2898 tracing_disabled = 1;
a98a3c3f 2899 }
3928a8a2 2900 goto out;
a98a3c3f 2901 }
3eefae99 2902
3928a8a2 2903 global_trace.entries = val;
a98a3c3f
SR
2904 }
2905
2906 filp->f_pos += cnt;
2907
19384c03
SR
2908 /* If check pages failed, return ENOMEM */
2909 if (tracing_disabled)
2910 cnt = -ENOMEM;
a98a3c3f 2911 out:
bf5e6519
SR
2912 for_each_tracing_cpu(cpu) {
2913 if (global_trace.data[cpu])
2914 atomic_dec(&global_trace.data[cpu]->disabled);
2915 if (max_tr.data[cpu])
2916 atomic_dec(&max_tr.data[cpu]->disabled);
2917 }
2918
c76f0694 2919 tracing_start();
a98a3c3f
SR
2920 max_tr.entries = global_trace.entries;
2921 mutex_unlock(&trace_types_lock);
2922
2923 return cnt;
2924}
2925
5bf9a1ee
PP
2926static int mark_printk(const char *fmt, ...)
2927{
2928 int ret;
2929 va_list args;
2930 va_start(args, fmt);
1fd8f2a3 2931 ret = trace_vprintk(0, -1, fmt, args);
5bf9a1ee
PP
2932 va_end(args);
2933 return ret;
2934}
2935
2936static ssize_t
2937tracing_mark_write(struct file *filp, const char __user *ubuf,
2938 size_t cnt, loff_t *fpos)
2939{
2940 char *buf;
2941 char *end;
5bf9a1ee 2942
c76f0694 2943 if (tracing_disabled)
5bf9a1ee
PP
2944 return -EINVAL;
2945
2946 if (cnt > TRACE_BUF_SIZE)
2947 cnt = TRACE_BUF_SIZE;
2948
2949 buf = kmalloc(cnt + 1, GFP_KERNEL);
2950 if (buf == NULL)
2951 return -ENOMEM;
2952
2953 if (copy_from_user(buf, ubuf, cnt)) {
2954 kfree(buf);
2955 return -EFAULT;
2956 }
2957
2958 /* Cut from the first nil or newline. */
2959 buf[cnt] = '\0';
2960 end = strchr(buf, '\n');
2961 if (end)
2962 *end = '\0';
2963
2964 cnt = mark_printk("%s\n", buf);
2965 kfree(buf);
2966 *fpos += cnt;
2967
2968 return cnt;
2969}
2970
bc0c38d1 2971static struct file_operations tracing_max_lat_fops = {
4bf39a94
IM
2972 .open = tracing_open_generic,
2973 .read = tracing_max_lat_read,
2974 .write = tracing_max_lat_write,
bc0c38d1
SR
2975};
2976
2977static struct file_operations tracing_ctrl_fops = {
4bf39a94
IM
2978 .open = tracing_open_generic,
2979 .read = tracing_ctrl_read,
2980 .write = tracing_ctrl_write,
bc0c38d1
SR
2981};
2982
2983static struct file_operations set_tracer_fops = {
4bf39a94
IM
2984 .open = tracing_open_generic,
2985 .read = tracing_set_trace_read,
2986 .write = tracing_set_trace_write,
bc0c38d1
SR
2987};
2988
b3806b43 2989static struct file_operations tracing_pipe_fops = {
4bf39a94 2990 .open = tracing_open_pipe,
2a2cc8f7 2991 .poll = tracing_poll_pipe,
4bf39a94 2992 .read = tracing_read_pipe,
3c56819b 2993 .splice_read = tracing_splice_read_pipe,
4bf39a94 2994 .release = tracing_release_pipe,
b3806b43
SR
2995};
2996
a98a3c3f
SR
2997static struct file_operations tracing_entries_fops = {
2998 .open = tracing_open_generic,
2999 .read = tracing_entries_read,
3000 .write = tracing_entries_write,
3001};
3002
5bf9a1ee 3003static struct file_operations tracing_mark_fops = {
43a15386 3004 .open = tracing_open_generic,
5bf9a1ee
PP
3005 .write = tracing_mark_write,
3006};
3007
bc0c38d1
SR
3008#ifdef CONFIG_DYNAMIC_FTRACE
3009
b807c3d0
SR
3010int __weak ftrace_arch_read_dyn_info(char *buf, int size)
3011{
3012 return 0;
3013}
3014
bc0c38d1 3015static ssize_t
b807c3d0 3016tracing_read_dyn_info(struct file *filp, char __user *ubuf,
bc0c38d1
SR
3017 size_t cnt, loff_t *ppos)
3018{
a26a2a27
SR
3019 static char ftrace_dyn_info_buffer[1024];
3020 static DEFINE_MUTEX(dyn_info_mutex);
bc0c38d1 3021 unsigned long *p = filp->private_data;
b807c3d0 3022 char *buf = ftrace_dyn_info_buffer;
a26a2a27 3023 int size = ARRAY_SIZE(ftrace_dyn_info_buffer);
bc0c38d1
SR
3024 int r;
3025
b807c3d0
SR
3026 mutex_lock(&dyn_info_mutex);
3027 r = sprintf(buf, "%ld ", *p);
4bf39a94 3028
a26a2a27 3029 r += ftrace_arch_read_dyn_info(buf+r, (size-1)-r);
b807c3d0
SR
3030 buf[r++] = '\n';
3031
3032 r = simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
3033
3034 mutex_unlock(&dyn_info_mutex);
3035
3036 return r;
bc0c38d1
SR
3037}
3038
b807c3d0 3039static struct file_operations tracing_dyn_info_fops = {
4bf39a94 3040 .open = tracing_open_generic,
b807c3d0 3041 .read = tracing_read_dyn_info,
bc0c38d1
SR
3042};
3043#endif
3044
3045static struct dentry *d_tracer;
3046
3047struct dentry *tracing_init_dentry(void)
3048{
3049 static int once;
3050
3051 if (d_tracer)
3052 return d_tracer;
3053
3054 d_tracer = debugfs_create_dir("tracing", NULL);
3055
3056 if (!d_tracer && !once) {
3057 once = 1;
3058 pr_warning("Could not create debugfs directory 'tracing'\n");
3059 return NULL;
3060 }
3061
3062 return d_tracer;
3063}
3064
b04cc6b1
FW
3065static struct dentry *d_percpu;
3066
3067struct dentry *tracing_dentry_percpu(void)
3068{
3069 static int once;
3070 struct dentry *d_tracer;
3071
3072 if (d_percpu)
3073 return d_percpu;
3074
3075 d_tracer = tracing_init_dentry();
3076
3077 if (!d_tracer)
3078 return NULL;
3079
3080 d_percpu = debugfs_create_dir("per_cpu", d_tracer);
3081
3082 if (!d_percpu && !once) {
3083 once = 1;
3084 pr_warning("Could not create debugfs directory 'per_cpu'\n");
3085 return NULL;
3086 }
3087
3088 return d_percpu;
3089}
3090
3091static void tracing_init_debugfs_percpu(long cpu)
3092{
3093 struct dentry *d_percpu = tracing_dentry_percpu();
8656e7a2
FW
3094 struct dentry *entry, *d_cpu;
3095 /* strlen(cpu) + MAX(log10(cpu)) + '\0' */
3096 char cpu_dir[7];
b04cc6b1
FW
3097
3098 if (cpu > 999 || cpu < 0)
3099 return;
3100
8656e7a2
FW
3101 sprintf(cpu_dir, "cpu%ld", cpu);
3102 d_cpu = debugfs_create_dir(cpu_dir, d_percpu);
3103 if (!d_cpu) {
3104 pr_warning("Could not create debugfs '%s' entry\n", cpu_dir);
3105 return;
3106 }
b04cc6b1 3107
8656e7a2
FW
3108 /* per cpu trace_pipe */
3109 entry = debugfs_create_file("trace_pipe", 0444, d_cpu,
b04cc6b1
FW
3110 (void *) cpu, &tracing_pipe_fops);
3111 if (!entry)
8656e7a2 3112 pr_warning("Could not create debugfs 'trace_pipe' entry\n");
b04cc6b1
FW
3113
3114 /* per cpu trace */
8656e7a2 3115 entry = debugfs_create_file("trace", 0444, d_cpu,
b04cc6b1
FW
3116 (void *) cpu, &tracing_fops);
3117 if (!entry)
8656e7a2 3118 pr_warning("Could not create debugfs 'trace' entry\n");
b04cc6b1
FW
3119}
3120
60a11774
SR
3121#ifdef CONFIG_FTRACE_SELFTEST
3122/* Let selftest have access to static functions in this file */
3123#include "trace_selftest.c"
3124#endif
3125
577b785f
SR
3126struct trace_option_dentry {
3127 struct tracer_opt *opt;
3128 struct tracer_flags *flags;
3129 struct dentry *entry;
3130};
3131
3132static ssize_t
3133trace_options_read(struct file *filp, char __user *ubuf, size_t cnt,
3134 loff_t *ppos)
3135{
3136 struct trace_option_dentry *topt = filp->private_data;
3137 char *buf;
3138
3139 if (topt->flags->val & topt->opt->bit)
3140 buf = "1\n";
3141 else
3142 buf = "0\n";
3143
3144 return simple_read_from_buffer(ubuf, cnt, ppos, buf, 2);
3145}
3146
3147static ssize_t
3148trace_options_write(struct file *filp, const char __user *ubuf, size_t cnt,
3149 loff_t *ppos)
3150{
3151 struct trace_option_dentry *topt = filp->private_data;
3152 unsigned long val;
3153 char buf[64];
3154 int ret;
3155
3156 if (cnt >= sizeof(buf))
3157 return -EINVAL;
3158
3159 if (copy_from_user(&buf, ubuf, cnt))
3160 return -EFAULT;
3161
3162 buf[cnt] = 0;
3163
3164 ret = strict_strtoul(buf, 10, &val);
3165 if (ret < 0)
3166 return ret;
3167
3168 ret = 0;
3169 switch (val) {
3170 case 0:
3171 /* do nothing if already cleared */
3172 if (!(topt->flags->val & topt->opt->bit))
3173 break;
3174
3175 mutex_lock(&trace_types_lock);
3176 if (current_trace->set_flag)
3177 ret = current_trace->set_flag(topt->flags->val,
3178 topt->opt->bit, 0);
3179 mutex_unlock(&trace_types_lock);
3180 if (ret)
3181 return ret;
3182 topt->flags->val &= ~topt->opt->bit;
3183 break;
3184 case 1:
3185 /* do nothing if already set */
3186 if (topt->flags->val & topt->opt->bit)
3187 break;
3188
3189 mutex_lock(&trace_types_lock);
3190 if (current_trace->set_flag)
3191 ret = current_trace->set_flag(topt->flags->val,
3192 topt->opt->bit, 1);
3193 mutex_unlock(&trace_types_lock);
3194 if (ret)
3195 return ret;
3196 topt->flags->val |= topt->opt->bit;
3197 break;
3198
3199 default:
3200 return -EINVAL;
3201 }
3202
3203 *ppos += cnt;
3204
3205 return cnt;
3206}
3207
3208
3209static const struct file_operations trace_options_fops = {
3210 .open = tracing_open_generic,
3211 .read = trace_options_read,
3212 .write = trace_options_write,
3213};
3214
a8259075
SR
3215static ssize_t
3216trace_options_core_read(struct file *filp, char __user *ubuf, size_t cnt,
3217 loff_t *ppos)
3218{
3219 long index = (long)filp->private_data;
3220 char *buf;
3221
3222 if (trace_flags & (1 << index))
3223 buf = "1\n";
3224 else
3225 buf = "0\n";
3226
3227 return simple_read_from_buffer(ubuf, cnt, ppos, buf, 2);
3228}
3229
3230static ssize_t
3231trace_options_core_write(struct file *filp, const char __user *ubuf, size_t cnt,
3232 loff_t *ppos)
3233{
3234 long index = (long)filp->private_data;
3235 char buf[64];
3236 unsigned long val;
3237 int ret;
3238
3239 if (cnt >= sizeof(buf))
3240 return -EINVAL;
3241
3242 if (copy_from_user(&buf, ubuf, cnt))
3243 return -EFAULT;
3244
3245 buf[cnt] = 0;
3246
3247 ret = strict_strtoul(buf, 10, &val);
3248 if (ret < 0)
3249 return ret;
3250
3251 switch (val) {
3252 case 0:
3253 trace_flags &= ~(1 << index);
3254 break;
3255 case 1:
3256 trace_flags |= 1 << index;
3257 break;
3258
3259 default:
3260 return -EINVAL;
3261 }
3262
3263 *ppos += cnt;
3264
3265 return cnt;
3266}
3267
a8259075
SR
3268static const struct file_operations trace_options_core_fops = {
3269 .open = tracing_open_generic,
3270 .read = trace_options_core_read,
3271 .write = trace_options_core_write,
3272};
3273
3274static struct dentry *trace_options_init_dentry(void)
3275{
3276 struct dentry *d_tracer;
3277 static struct dentry *t_options;
3278
3279 if (t_options)
3280 return t_options;
3281
3282 d_tracer = tracing_init_dentry();
3283 if (!d_tracer)
3284 return NULL;
3285
3286 t_options = debugfs_create_dir("options", d_tracer);
3287 if (!t_options) {
3288 pr_warning("Could not create debugfs directory 'options'\n");
3289 return NULL;
3290 }
3291
3292 return t_options;
3293}
3294
577b785f
SR
3295static void
3296create_trace_option_file(struct trace_option_dentry *topt,
3297 struct tracer_flags *flags,
3298 struct tracer_opt *opt)
3299{
3300 struct dentry *t_options;
3301 struct dentry *entry;
3302
3303 t_options = trace_options_init_dentry();
3304 if (!t_options)
3305 return;
3306
3307 topt->flags = flags;
3308 topt->opt = opt;
3309
3310 entry = debugfs_create_file(opt->name, 0644, t_options, topt,
3311 &trace_options_fops);
3312
3313 topt->entry = entry;
3314
3315}
3316
3317static struct trace_option_dentry *
3318create_trace_option_files(struct tracer *tracer)
3319{
3320 struct trace_option_dentry *topts;
3321 struct tracer_flags *flags;
3322 struct tracer_opt *opts;
3323 int cnt;
3324
3325 if (!tracer)
3326 return NULL;
3327
3328 flags = tracer->flags;
3329
3330 if (!flags || !flags->opts)
3331 return NULL;
3332
3333 opts = flags->opts;
3334
3335 for (cnt = 0; opts[cnt].name; cnt++)
3336 ;
3337
0cfe8245 3338 topts = kcalloc(cnt + 1, sizeof(*topts), GFP_KERNEL);
577b785f
SR
3339 if (!topts)
3340 return NULL;
3341
3342 for (cnt = 0; opts[cnt].name; cnt++)
3343 create_trace_option_file(&topts[cnt], flags,
3344 &opts[cnt]);
3345
3346 return topts;
3347}
3348
3349static void
3350destroy_trace_option_files(struct trace_option_dentry *topts)
3351{
3352 int cnt;
3353
3354 if (!topts)
3355 return;
3356
3357 for (cnt = 0; topts[cnt].opt; cnt++) {
3358 if (topts[cnt].entry)
3359 debugfs_remove(topts[cnt].entry);
3360 }
3361
3362 kfree(topts);
3363}
3364
a8259075
SR
3365static struct dentry *
3366create_trace_option_core_file(const char *option, long index)
3367{
3368 struct dentry *t_options;
3369 struct dentry *entry;
3370
3371 t_options = trace_options_init_dentry();
3372 if (!t_options)
3373 return NULL;
3374
3375 entry = debugfs_create_file(option, 0644, t_options, (void *)index,
3376 &trace_options_core_fops);
3377
3378 return entry;
3379}
3380
3381static __init void create_trace_options_dir(void)
3382{
3383 struct dentry *t_options;
3384 struct dentry *entry;
3385 int i;
3386
3387 t_options = trace_options_init_dentry();
3388 if (!t_options)
3389 return;
3390
3391 for (i = 0; trace_options[i]; i++) {
3392 entry = create_trace_option_core_file(trace_options[i], i);
3393 if (!entry)
3394 pr_warning("Could not create debugfs %s entry\n",
3395 trace_options[i]);
3396 }
3397}
3398
b5ad384e 3399static __init int tracer_init_debugfs(void)
bc0c38d1
SR
3400{
3401 struct dentry *d_tracer;
3402 struct dentry *entry;
b04cc6b1 3403 int cpu;
bc0c38d1
SR
3404
3405 d_tracer = tracing_init_dentry();
3406
3407 entry = debugfs_create_file("tracing_enabled", 0644, d_tracer,
3408 &global_trace, &tracing_ctrl_fops);
3409 if (!entry)
3410 pr_warning("Could not create debugfs 'tracing_enabled' entry\n");
3411
ee6bce52 3412 entry = debugfs_create_file("trace_options", 0644, d_tracer,
bc0c38d1
SR
3413 NULL, &tracing_iter_fops);
3414 if (!entry)
ee6bce52 3415 pr_warning("Could not create debugfs 'trace_options' entry\n");
bc0c38d1 3416
a8259075
SR
3417 create_trace_options_dir();
3418
c7078de1
IM
3419 entry = debugfs_create_file("tracing_cpumask", 0644, d_tracer,
3420 NULL, &tracing_cpumask_fops);
3421 if (!entry)
3422 pr_warning("Could not create debugfs 'tracing_cpumask' entry\n");
3423
bc0c38d1 3424 entry = debugfs_create_file("trace", 0444, d_tracer,
b04cc6b1 3425 (void *) TRACE_PIPE_ALL_CPU, &tracing_fops);
bc0c38d1
SR
3426 if (!entry)
3427 pr_warning("Could not create debugfs 'trace' entry\n");
3428
3429 entry = debugfs_create_file("available_tracers", 0444, d_tracer,
3430 &global_trace, &show_traces_fops);
3431 if (!entry)
98a983aa 3432 pr_warning("Could not create debugfs 'available_tracers' entry\n");
bc0c38d1
SR
3433
3434 entry = debugfs_create_file("current_tracer", 0444, d_tracer,
3435 &global_trace, &set_tracer_fops);
3436 if (!entry)
98a983aa 3437 pr_warning("Could not create debugfs 'current_tracer' entry\n");
bc0c38d1
SR
3438
3439 entry = debugfs_create_file("tracing_max_latency", 0644, d_tracer,
3440 &tracing_max_latency,
3441 &tracing_max_lat_fops);
3442 if (!entry)
3443 pr_warning("Could not create debugfs "
3444 "'tracing_max_latency' entry\n");
3445
3446 entry = debugfs_create_file("tracing_thresh", 0644, d_tracer,
3447 &tracing_thresh, &tracing_max_lat_fops);
3448 if (!entry)
3449 pr_warning("Could not create debugfs "
98a983aa 3450 "'tracing_thresh' entry\n");
7bd2f24c
IM
3451 entry = debugfs_create_file("README", 0644, d_tracer,
3452 NULL, &tracing_readme_fops);
3453 if (!entry)
3454 pr_warning("Could not create debugfs 'README' entry\n");
3455
b04cc6b1
FW
3456 entry = debugfs_create_file("trace_pipe", 0444, d_tracer,
3457 (void *) TRACE_PIPE_ALL_CPU, &tracing_pipe_fops);
b3806b43
SR
3458 if (!entry)
3459 pr_warning("Could not create debugfs "
98a983aa 3460 "'trace_pipe' entry\n");
bc0c38d1 3461
a94c80e7 3462 entry = debugfs_create_file("buffer_size_kb", 0644, d_tracer,
a98a3c3f
SR
3463 &global_trace, &tracing_entries_fops);
3464 if (!entry)
3465 pr_warning("Could not create debugfs "
a94c80e7 3466 "'buffer_size_kb' entry\n");
a98a3c3f 3467
5bf9a1ee
PP
3468 entry = debugfs_create_file("trace_marker", 0220, d_tracer,
3469 NULL, &tracing_mark_fops);
3470 if (!entry)
3471 pr_warning("Could not create debugfs "
3472 "'trace_marker' entry\n");
3473
bc0c38d1
SR
3474#ifdef CONFIG_DYNAMIC_FTRACE
3475 entry = debugfs_create_file("dyn_ftrace_total_info", 0444, d_tracer,
3476 &ftrace_update_tot_cnt,
b807c3d0 3477 &tracing_dyn_info_fops);
bc0c38d1
SR
3478 if (!entry)
3479 pr_warning("Could not create debugfs "
3480 "'dyn_ftrace_total_info' entry\n");
3481#endif
d618b3e6
IM
3482#ifdef CONFIG_SYSPROF_TRACER
3483 init_tracer_sysprof_debugfs(d_tracer);
3484#endif
b04cc6b1
FW
3485
3486 for_each_tracing_cpu(cpu)
3487 tracing_init_debugfs_percpu(cpu);
3488
b5ad384e 3489 return 0;
bc0c38d1
SR
3490}
3491
1fd8f2a3 3492int trace_vprintk(unsigned long ip, int depth, const char *fmt, va_list args)
dd0e545f 3493{
dd0e545f
SR
3494 static DEFINE_SPINLOCK(trace_buf_lock);
3495 static char trace_buf[TRACE_BUF_SIZE];
f09ce573 3496
3928a8a2 3497 struct ring_buffer_event *event;
f09ce573 3498 struct trace_array *tr = &global_trace;
dd0e545f 3499 struct trace_array_cpu *data;
38697053 3500 int cpu, len = 0, size, pc;
e726f5f9
IM
3501 struct print_entry *entry;
3502 unsigned long irq_flags;
dd0e545f 3503
8e1b82e0 3504 if (tracing_disabled || tracing_selftest_running)
dd0e545f
SR
3505 return 0;
3506
38697053
SR
3507 pc = preempt_count();
3508 preempt_disable_notrace();
dd0e545f
SR
3509 cpu = raw_smp_processor_id();
3510 data = tr->data[cpu];
dd0e545f 3511
3ea2e6d7 3512 if (unlikely(atomic_read(&data->disabled)))
dd0e545f
SR
3513 goto out;
3514
380c4b14
FW
3515 pause_graph_tracing();
3516 spin_lock_irqsave(&trace_buf_lock, irq_flags);
801fe400 3517 len = vsnprintf(trace_buf, TRACE_BUF_SIZE, fmt, args);
dd0e545f
SR
3518
3519 len = min(len, TRACE_BUF_SIZE-1);
3520 trace_buf[len] = 0;
3521
777e208d 3522 size = sizeof(*entry) + len + 1;
51a763dd 3523 event = trace_buffer_lock_reserve(tr, TRACE_PRINT, size, irq_flags, pc);
3928a8a2
SR
3524 if (!event)
3525 goto out_unlock;
777e208d 3526 entry = ring_buffer_event_data(event);
777e208d 3527 entry->ip = ip;
1fd8f2a3 3528 entry->depth = depth;
dd0e545f 3529
777e208d
SR
3530 memcpy(&entry->buf, trace_buf, len);
3531 entry->buf[len] = 0;
0a987751 3532 ring_buffer_unlock_commit(tr->buffer, event);
dd0e545f 3533
3928a8a2 3534 out_unlock:
380c4b14
FW
3535 spin_unlock_irqrestore(&trace_buf_lock, irq_flags);
3536 unpause_graph_tracing();
dd0e545f 3537 out:
38697053 3538 preempt_enable_notrace();
dd0e545f
SR
3539
3540 return len;
3541}
801fe400
PP
3542EXPORT_SYMBOL_GPL(trace_vprintk);
3543
3544int __ftrace_printk(unsigned long ip, const char *fmt, ...)
3545{
3546 int ret;
3547 va_list ap;
3548
3549 if (!(trace_flags & TRACE_ITER_PRINTK))
3550 return 0;
3551
3552 va_start(ap, fmt);
21a8c466 3553 ret = trace_vprintk(ip, task_curr_ret_stack(current), fmt, ap);
801fe400
PP
3554 va_end(ap);
3555 return ret;
3556}
dd0e545f
SR
3557EXPORT_SYMBOL_GPL(__ftrace_printk);
3558
9011262a
ACM
3559int __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap)
3560{
3561 if (!(trace_flags & TRACE_ITER_PRINTK))
3562 return 0;
3563
3564 return trace_vprintk(ip, task_curr_ret_stack(current), fmt, ap);
3565}
3566EXPORT_SYMBOL_GPL(__ftrace_vprintk);
3567
3f5a54e3
SR
3568static int trace_panic_handler(struct notifier_block *this,
3569 unsigned long event, void *unused)
3570{
944ac425
SR
3571 if (ftrace_dump_on_oops)
3572 ftrace_dump();
3f5a54e3
SR
3573 return NOTIFY_OK;
3574}
3575
3576static struct notifier_block trace_panic_notifier = {
3577 .notifier_call = trace_panic_handler,
3578 .next = NULL,
3579 .priority = 150 /* priority: INT_MAX >= x >= 0 */
3580};
3581
3582static int trace_die_handler(struct notifier_block *self,
3583 unsigned long val,
3584 void *data)
3585{
3586 switch (val) {
3587 case DIE_OOPS:
944ac425
SR
3588 if (ftrace_dump_on_oops)
3589 ftrace_dump();
3f5a54e3
SR
3590 break;
3591 default:
3592 break;
3593 }
3594 return NOTIFY_OK;
3595}
3596
3597static struct notifier_block trace_die_notifier = {
3598 .notifier_call = trace_die_handler,
3599 .priority = 200
3600};
3601
3602/*
3603 * printk is set to max of 1024, we really don't need it that big.
3604 * Nothing should be printing 1000 characters anyway.
3605 */
3606#define TRACE_MAX_PRINT 1000
3607
3608/*
3609 * Define here KERN_TRACE so that we have one place to modify
3610 * it if we decide to change what log level the ftrace dump
3611 * should be at.
3612 */
428aee14 3613#define KERN_TRACE KERN_EMERG
3f5a54e3
SR
3614
3615static void
3616trace_printk_seq(struct trace_seq *s)
3617{
3618 /* Probably should print a warning here. */
3619 if (s->len >= 1000)
3620 s->len = 1000;
3621
3622 /* should be zero ended, but we are paranoid. */
3623 s->buffer[s->len] = 0;
3624
3625 printk(KERN_TRACE "%s", s->buffer);
3626
f9520750 3627 trace_seq_init(s);
3f5a54e3
SR
3628}
3629
3f5a54e3
SR
3630void ftrace_dump(void)
3631{
3632 static DEFINE_SPINLOCK(ftrace_dump_lock);
3633 /* use static because iter can be a bit big for the stack */
3634 static struct trace_iterator iter;
3f5a54e3 3635 static int dump_ran;
d769041f
SR
3636 unsigned long flags;
3637 int cnt = 0, cpu;
3f5a54e3
SR
3638
3639 /* only one dump */
3640 spin_lock_irqsave(&ftrace_dump_lock, flags);
3641 if (dump_ran)
3642 goto out;
3643
3644 dump_ran = 1;
3645
3646 /* No turning back! */
0ee6b6cf 3647 tracing_off();
81adbdc0 3648 ftrace_kill();
3f5a54e3 3649
d769041f
SR
3650 for_each_tracing_cpu(cpu) {
3651 atomic_inc(&global_trace.data[cpu]->disabled);
3652 }
3653
b54d3de9
TE
3654 /* don't look at user memory in panic mode */
3655 trace_flags &= ~TRACE_ITER_SYM_USEROBJ;
3656
3f5a54e3
SR
3657 printk(KERN_TRACE "Dumping ftrace buffer:\n");
3658
3659 iter.tr = &global_trace;
3660 iter.trace = current_trace;
3661
3662 /*
3663 * We need to stop all tracing on all CPUS to read the
3664 * the next buffer. This is a bit expensive, but is
3665 * not done often. We fill all what we can read,
3666 * and then release the locks again.
3667 */
3668
3f5a54e3
SR
3669 while (!trace_empty(&iter)) {
3670
3671 if (!cnt)
3672 printk(KERN_TRACE "---------------------------------\n");
3673
3674 cnt++;
3675
3676 /* reset all but tr, trace, and overruns */
3677 memset(&iter.seq, 0,
3678 sizeof(struct trace_iterator) -
3679 offsetof(struct trace_iterator, seq));
3680 iter.iter_flags |= TRACE_FILE_LAT_FMT;
3681 iter.pos = -1;
3682
3683 if (find_next_entry_inc(&iter) != NULL) {
3684 print_trace_line(&iter);
3685 trace_consume(&iter);
3686 }
3687
3688 trace_printk_seq(&iter.seq);
3689 }
3690
3691 if (!cnt)
3692 printk(KERN_TRACE " (ftrace buffer empty)\n");
3693 else
3694 printk(KERN_TRACE "---------------------------------\n");
3695
3f5a54e3
SR
3696 out:
3697 spin_unlock_irqrestore(&ftrace_dump_lock, flags);
3698}
3699
3928a8a2 3700__init static int tracer_alloc_buffers(void)
bc0c38d1 3701{
4c11d7ae 3702 struct trace_array_cpu *data;
4c11d7ae 3703 int i;
9e01c1b7 3704 int ret = -ENOMEM;
4c11d7ae 3705
9e01c1b7
RR
3706 if (!alloc_cpumask_var(&tracing_buffer_mask, GFP_KERNEL))
3707 goto out;
3708
3709 if (!alloc_cpumask_var(&tracing_cpumask, GFP_KERNEL))
3710 goto out_free_buffer_mask;
4c11d7ae 3711
b04cc6b1
FW
3712 if (!alloc_cpumask_var(&tracing_reader_cpumask, GFP_KERNEL))
3713 goto out_free_tracing_cpumask;
3714
9e01c1b7
RR
3715 cpumask_copy(tracing_buffer_mask, cpu_possible_mask);
3716 cpumask_copy(tracing_cpumask, cpu_all_mask);
b04cc6b1 3717 cpumask_clear(tracing_reader_cpumask);
9e01c1b7
RR
3718
3719 /* TODO: make the number of buffers hot pluggable with CPUS */
3928a8a2
SR
3720 global_trace.buffer = ring_buffer_alloc(trace_buf_size,
3721 TRACE_BUFFER_FLAGS);
3722 if (!global_trace.buffer) {
3723 printk(KERN_ERR "tracer: failed to allocate ring buffer!\n");
3724 WARN_ON(1);
9e01c1b7 3725 goto out_free_cpumask;
4c11d7ae 3726 }
3928a8a2 3727 global_trace.entries = ring_buffer_size(global_trace.buffer);
4c11d7ae 3728
9e01c1b7 3729
4c11d7ae 3730#ifdef CONFIG_TRACER_MAX_TRACE
3928a8a2
SR
3731 max_tr.buffer = ring_buffer_alloc(trace_buf_size,
3732 TRACE_BUFFER_FLAGS);
3733 if (!max_tr.buffer) {
3734 printk(KERN_ERR "tracer: failed to allocate max ring buffer!\n");
3735 WARN_ON(1);
3736 ring_buffer_free(global_trace.buffer);
9e01c1b7 3737 goto out_free_cpumask;
4c11d7ae 3738 }
3928a8a2
SR
3739 max_tr.entries = ring_buffer_size(max_tr.buffer);
3740 WARN_ON(max_tr.entries != global_trace.entries);
a98a3c3f 3741#endif
ab46428c 3742
4c11d7ae 3743 /* Allocate the first page for all buffers */
ab46428c 3744 for_each_tracing_cpu(i) {
4c11d7ae 3745 data = global_trace.data[i] = &per_cpu(global_trace_cpu, i);
bc0c38d1 3746 max_tr.data[i] = &per_cpu(max_data, i);
4c11d7ae 3747 }
bc0c38d1 3748
bc0c38d1
SR
3749 trace_init_cmdlines();
3750
43a15386 3751 register_tracer(&nop_trace);
79fb0768 3752 current_trace = &nop_trace;
b5ad384e
FW
3753#ifdef CONFIG_BOOT_TRACER
3754 register_tracer(&boot_tracer);
b5ad384e 3755#endif
60a11774
SR
3756 /* All seems OK, enable tracing */
3757 tracing_disabled = 0;
3928a8a2 3758
3f5a54e3
SR
3759 atomic_notifier_chain_register(&panic_notifier_list,
3760 &trace_panic_notifier);
3761
3762 register_die_notifier(&trace_die_notifier);
9e01c1b7 3763 ret = 0;
3f5a54e3 3764
9e01c1b7 3765out_free_cpumask:
b04cc6b1
FW
3766 free_cpumask_var(tracing_reader_cpumask);
3767out_free_tracing_cpumask:
9e01c1b7
RR
3768 free_cpumask_var(tracing_cpumask);
3769out_free_buffer_mask:
3770 free_cpumask_var(tracing_buffer_mask);
3771out:
3772 return ret;
bc0c38d1 3773}
b2821ae6
SR
3774
3775__init static int clear_boot_tracer(void)
3776{
3777 /*
3778 * The default tracer at boot buffer is an init section.
3779 * This function is called in lateinit. If we did not
3780 * find the boot tracer, then clear it out, to prevent
3781 * later registration from accessing the buffer that is
3782 * about to be freed.
3783 */
3784 if (!default_bootup_tracer)
3785 return 0;
3786
3787 printk(KERN_INFO "ftrace bootup tracer '%s' not registered.\n",
3788 default_bootup_tracer);
3789 default_bootup_tracer = NULL;
3790
3791 return 0;
3792}
3793
b5ad384e
FW
3794early_initcall(tracer_alloc_buffers);
3795fs_initcall(tracer_init_debugfs);
b2821ae6 3796late_initcall(clear_boot_tracer);